{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"\n",
" $\n",
" \\DeclareMathOperator*{\\argmin}{arg\\,min} \n",
" \\DeclareMathOperator*{\\argmax}{arg\\,max}\n",
" \\DeclareMathOperator*{\\median}{median}\n",
" \\newcommand{\\dydx}[2]{\\frac{\\partial #1}{\\partial #2}}\n",
" \\newcommand{\\x}{\\times}\n",
" \\newcommand{\\CWT}{\\mathbb{C}\\mathrm{WT}}\n",
" \\newcommand{\\DTCWT}{\\mathrm{DT}\\CWT}\n",
" \\newcommand{\\mat}[1]{\\mathbf{#1}}\n",
" $\n",
"
"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# Learnable Soft Shrinkage Thresholds\n",
"In this work, we want to extend the soft-thresholding wavelet ideas initially introduced by Donoho and Johnstone in [Ideal Spatial Adaptation by Wavelet Shrinkage](http://statweb.stanford.edu/~imj/WEBLIST/1994/isaws.pdf), and later developed on by Chang, Yu and Vetterli in [Adaptive Wavelet Thresholding for Image Denoising and Compression](https://ieeexplore.ieee.org/document/862633).\n",
"\n",
"In particular, given a noisy image and its clean version as a target, is it possible to learn via backpropagation the soft shrinkage thresholds? How much better are they than by using the estimated thresholds from Chang et. al?\n",
"\n",
"Note that this is a toy problem - I will be using the clean image as a target and use the MSE to backpropagate values to the thresholds. In general denoising problems we of course do not have access to the clean image. Perhaps we can estimate noise as a next step, but we will leave this for future work."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Background\n",
"Soft thresholding is a very popular and effective technique for denoising/compressing images. The basic technique involves:\n",
"\n",
"- Taking a wavelet transform of the input - this has the advantage that the wavelet coefficients for most subbands of natural images are quite sparse.\n",
"\n",
"\n",
"- Calculate a threshold $T$ that will convert the noisy image $Y=X+\\epsilon$ to a denoised estimate $\\hat{X}$. Do this by minimizing the Bayes Risk \n",
"$$\n",
"r(T) = E[(\\hat{X}(T) - X)^2]\n",
"$$ \n",
"\n",
"given some priors put on $p(\\epsilon)$ (a common one being that $\\epsilon \\sim N(0, \\sigma^2)$\n",
"- Use these thresholds on the wavelet bandpass coefficients (everything except from the LL output)\n",
"$$\n",
"\\eta(x) = sgn(x) max(|x| -T, 0)\n",
"$$\n",
"\n",
"\n",
"\n",
"- Reconstruct $\\hat{X}$ from the newly shrunk coefficients."
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {
"ExecuteTime": {
"end_time": "2018-12-03T16:51:21.617883Z",
"start_time": "2018-12-03T16:51:20.464673Z"
}
},
"outputs": [],
"source": [
"# Import some plotting libraries\n",
"%matplotlib notebook\n",
"import matplotlib.pyplot as plt\n",
"import matplotlib.gridspec as gridspec\n",
"from matplotlib.ticker import FormatStrFormatter\n",
"from PIL import Image\n",
"import plotters # Can be obtained from https://github.com/fbcotter/plotters\n",
"\n",
"# Import our numeric libraries\n",
"import numpy as np\n",
"import torch\n",
"import torch.nn as nn\n",
"\n",
"# import our wavelet libraries. pytorch_wavelets can be obtained from\n",
"# https://github.com/fbcotter/pytorch_wavelets\n",
"import pywt\n",
"from pytorch_wavelets import DTCWTForward, DTCWTInverse, DWTForward, DWTInverse"
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {
"ExecuteTime": {
"end_time": "2018-12-03T16:51:22.002931Z",
"start_time": "2018-12-03T16:51:21.619129Z"
}
},
"outputs": [
{
"data": {
"application/javascript": [
"/* Put everything inside the global mpl namespace */\n",
"window.mpl = {};\n",
"\n",
"\n",
"mpl.get_websocket_type = function() {\n",
" if (typeof(WebSocket) !== 'undefined') {\n",
" return WebSocket;\n",
" } else if (typeof(MozWebSocket) !== 'undefined') {\n",
" return MozWebSocket;\n",
" } else {\n",
" alert('Your browser does not have WebSocket support.' +\n",
" 'Please try Chrome, Safari or Firefox ≥ 6. ' +\n",
" 'Firefox 4 and 5 are also supported but you ' +\n",
" 'have to enable WebSockets in about:config.');\n",
" };\n",
"}\n",
"\n",
"mpl.figure = function(figure_id, websocket, ondownload, parent_element) {\n",
" this.id = figure_id;\n",
"\n",
" this.ws = websocket;\n",
"\n",
" this.supports_binary = (this.ws.binaryType != undefined);\n",
"\n",
" if (!this.supports_binary) {\n",
" var warnings = document.getElementById(\"mpl-warnings\");\n",
" if (warnings) {\n",
" warnings.style.display = 'block';\n",
" warnings.textContent = (\n",
" \"This browser does not support binary websocket messages. \" +\n",
" \"Performance may be slow.\");\n",
" }\n",
" }\n",
"\n",
" this.imageObj = new Image();\n",
"\n",
" this.context = undefined;\n",
" this.message = undefined;\n",
" this.canvas = undefined;\n",
" this.rubberband_canvas = undefined;\n",
" this.rubberband_context = undefined;\n",
" this.format_dropdown = undefined;\n",
"\n",
" this.image_mode = 'full';\n",
"\n",
" this.root = $('');\n",
" this._root_extra_style(this.root)\n",
" this.root.attr('style', 'display: inline-block');\n",
"\n",
" $(parent_element).append(this.root);\n",
"\n",
" this._init_header(this);\n",
" this._init_canvas(this);\n",
" this._init_toolbar(this);\n",
"\n",
" var fig = this;\n",
"\n",
" this.waiting = false;\n",
"\n",
" this.ws.onopen = function () {\n",
" fig.send_message(\"supports_binary\", {value: fig.supports_binary});\n",
" fig.send_message(\"send_image_mode\", {});\n",
" if (mpl.ratio != 1) {\n",
" fig.send_message(\"set_dpi_ratio\", {'dpi_ratio': mpl.ratio});\n",
" }\n",
" fig.send_message(\"refresh\", {});\n",
" }\n",
"\n",
" this.imageObj.onload = function() {\n",
" if (fig.image_mode == 'full') {\n",
" // Full images could contain transparency (where diff images\n",
" // almost always do), so we need to clear the canvas so that\n",
" // there is no ghosting.\n",
" fig.context.clearRect(0, 0, fig.canvas.width, fig.canvas.height);\n",
" }\n",
" fig.context.drawImage(fig.imageObj, 0, 0);\n",
" };\n",
"\n",
" this.imageObj.onunload = function() {\n",
" fig.ws.close();\n",
" }\n",
"\n",
" this.ws.onmessage = this._make_on_message_function(this);\n",
"\n",
" this.ondownload = ondownload;\n",
"}\n",
"\n",
"mpl.figure.prototype._init_header = function() {\n",
" var titlebar = $(\n",
" '');\n",
" var titletext = $(\n",
" '');\n",
" titlebar.append(titletext)\n",
" this.root.append(titlebar);\n",
" this.header = titletext[0];\n",
"}\n",
"\n",
"\n",
"\n",
"mpl.figure.prototype._canvas_extra_style = function(canvas_div) {\n",
"\n",
"}\n",
"\n",
"\n",
"mpl.figure.prototype._root_extra_style = function(canvas_div) {\n",
"\n",
"}\n",
"\n",
"mpl.figure.prototype._init_canvas = function() {\n",
" var fig = this;\n",
"\n",
" var canvas_div = $('');\n",
"\n",
" canvas_div.attr('style', 'position: relative; clear: both; outline: 0');\n",
"\n",
" function canvas_keyboard_event(event) {\n",
" return fig.key_event(event, event['data']);\n",
" }\n",
"\n",
" canvas_div.keydown('key_press', canvas_keyboard_event);\n",
" canvas_div.keyup('key_release', canvas_keyboard_event);\n",
" this.canvas_div = canvas_div\n",
" this._canvas_extra_style(canvas_div)\n",
" this.root.append(canvas_div);\n",
"\n",
" var canvas = $('');\n",
" canvas.addClass('mpl-canvas');\n",
" canvas.attr('style', \"left: 0; top: 0; z-index: 0; outline: 0\")\n",
"\n",
" this.canvas = canvas[0];\n",
" this.context = canvas[0].getContext(\"2d\");\n",
"\n",
" var backingStore = this.context.backingStorePixelRatio ||\n",
"\tthis.context.webkitBackingStorePixelRatio ||\n",
"\tthis.context.mozBackingStorePixelRatio ||\n",
"\tthis.context.msBackingStorePixelRatio ||\n",
"\tthis.context.oBackingStorePixelRatio ||\n",
"\tthis.context.backingStorePixelRatio || 1;\n",
"\n",
" mpl.ratio = (window.devicePixelRatio || 1) / backingStore;\n",
"\n",
" var rubberband = $('');\n",
" rubberband.attr('style', \"position: absolute; left: 0; top: 0; z-index: 1;\")\n",
"\n",
" var pass_mouse_events = true;\n",
"\n",
" canvas_div.resizable({\n",
" start: function(event, ui) {\n",
" pass_mouse_events = false;\n",
" },\n",
" resize: function(event, ui) {\n",
" fig.request_resize(ui.size.width, ui.size.height);\n",
" },\n",
" stop: function(event, ui) {\n",
" pass_mouse_events = true;\n",
" fig.request_resize(ui.size.width, ui.size.height);\n",
" },\n",
" });\n",
"\n",
" function mouse_event_fn(event) {\n",
" if (pass_mouse_events)\n",
" return fig.mouse_event(event, event['data']);\n",
" }\n",
"\n",
" rubberband.mousedown('button_press', mouse_event_fn);\n",
" rubberband.mouseup('button_release', mouse_event_fn);\n",
" // Throttle sequential mouse events to 1 every 20ms.\n",
" rubberband.mousemove('motion_notify', mouse_event_fn);\n",
"\n",
" rubberband.mouseenter('figure_enter', mouse_event_fn);\n",
" rubberband.mouseleave('figure_leave', mouse_event_fn);\n",
"\n",
" canvas_div.on(\"wheel\", function (event) {\n",
" event = event.originalEvent;\n",
" event['data'] = 'scroll'\n",
" if (event.deltaY < 0) {\n",
" event.step = 1;\n",
" } else {\n",
" event.step = -1;\n",
" }\n",
" mouse_event_fn(event);\n",
" });\n",
"\n",
" canvas_div.append(canvas);\n",
" canvas_div.append(rubberband);\n",
"\n",
" this.rubberband = rubberband;\n",
" this.rubberband_canvas = rubberband[0];\n",
" this.rubberband_context = rubberband[0].getContext(\"2d\");\n",
" this.rubberband_context.strokeStyle = \"#000000\";\n",
"\n",
" this._resize_canvas = function(width, height) {\n",
" // Keep the size of the canvas, canvas container, and rubber band\n",
" // canvas in synch.\n",
" canvas_div.css('width', width)\n",
" canvas_div.css('height', height)\n",
"\n",
" canvas.attr('width', width * mpl.ratio);\n",
" canvas.attr('height', height * mpl.ratio);\n",
" canvas.attr('style', 'width: ' + width + 'px; height: ' + height + 'px;');\n",
"\n",
" rubberband.attr('width', width);\n",
" rubberband.attr('height', height);\n",
" }\n",
"\n",
" // Set the figure to an initial 600x600px, this will subsequently be updated\n",
" // upon first draw.\n",
" this._resize_canvas(600, 600);\n",
"\n",
" // Disable right mouse context menu.\n",
" $(this.rubberband_canvas).bind(\"contextmenu\",function(e){\n",
" return false;\n",
" });\n",
"\n",
" function set_focus () {\n",
" canvas.focus();\n",
" canvas_div.focus();\n",
" }\n",
"\n",
" window.setTimeout(set_focus, 100);\n",
"}\n",
"\n",
"mpl.figure.prototype._init_toolbar = function() {\n",
" var fig = this;\n",
"\n",
" var nav_element = $('')\n",
" nav_element.attr('style', 'width: 100%');\n",
" this.root.append(nav_element);\n",
"\n",
" // Define a callback function for later on.\n",
" function toolbar_event(event) {\n",
" return fig.toolbar_button_onclick(event['data']);\n",
" }\n",
" function toolbar_mouse_event(event) {\n",
" return fig.toolbar_button_onmouseover(event['data']);\n",
" }\n",
"\n",
" for(var toolbar_ind in mpl.toolbar_items) {\n",
" var name = mpl.toolbar_items[toolbar_ind][0];\n",
" var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
" var image = mpl.toolbar_items[toolbar_ind][2];\n",
" var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
"\n",
" if (!name) {\n",
" // put a spacer in here.\n",
" continue;\n",
" }\n",
" var button = $('');\n",
" button.addClass('ui-button ui-widget ui-state-default ui-corner-all ' +\n",
" 'ui-button-icon-only');\n",
" button.attr('role', 'button');\n",
" button.attr('aria-disabled', 'false');\n",
" button.click(method_name, toolbar_event);\n",
" button.mouseover(tooltip, toolbar_mouse_event);\n",
"\n",
" var icon_img = $('');\n",
" icon_img.addClass('ui-button-icon-primary ui-icon');\n",
" icon_img.addClass(image);\n",
" icon_img.addClass('ui-corner-all');\n",
"\n",
" var tooltip_span = $('');\n",
" tooltip_span.addClass('ui-button-text');\n",
" tooltip_span.html(tooltip);\n",
"\n",
" button.append(icon_img);\n",
" button.append(tooltip_span);\n",
"\n",
" nav_element.append(button);\n",
" }\n",
"\n",
" var fmt_picker_span = $('');\n",
"\n",
" var fmt_picker = $('');\n",
" fmt_picker.addClass('mpl-toolbar-option ui-widget ui-widget-content');\n",
" fmt_picker_span.append(fmt_picker);\n",
" nav_element.append(fmt_picker_span);\n",
" this.format_dropdown = fmt_picker[0];\n",
"\n",
" for (var ind in mpl.extensions) {\n",
" var fmt = mpl.extensions[ind];\n",
" var option = $(\n",
" '', {selected: fmt === mpl.default_extension}).html(fmt);\n",
" fmt_picker.append(option)\n",
" }\n",
"\n",
" // Add hover states to the ui-buttons\n",
" $( \".ui-button\" ).hover(\n",
" function() { $(this).addClass(\"ui-state-hover\");},\n",
" function() { $(this).removeClass(\"ui-state-hover\");}\n",
" );\n",
"\n",
" var status_bar = $('');\n",
" nav_element.append(status_bar);\n",
" this.message = status_bar[0];\n",
"}\n",
"\n",
"mpl.figure.prototype.request_resize = function(x_pixels, y_pixels) {\n",
" // Request matplotlib to resize the figure. Matplotlib will then trigger a resize in the client,\n",
" // which will in turn request a refresh of the image.\n",
" this.send_message('resize', {'width': x_pixels, 'height': y_pixels});\n",
"}\n",
"\n",
"mpl.figure.prototype.send_message = function(type, properties) {\n",
" properties['type'] = type;\n",
" properties['figure_id'] = this.id;\n",
" this.ws.send(JSON.stringify(properties));\n",
"}\n",
"\n",
"mpl.figure.prototype.send_draw_message = function() {\n",
" if (!this.waiting) {\n",
" this.waiting = true;\n",
" this.ws.send(JSON.stringify({type: \"draw\", figure_id: this.id}));\n",
" }\n",
"}\n",
"\n",
"\n",
"mpl.figure.prototype.handle_save = function(fig, msg) {\n",
" var format_dropdown = fig.format_dropdown;\n",
" var format = format_dropdown.options[format_dropdown.selectedIndex].value;\n",
" fig.ondownload(fig, format);\n",
"}\n",
"\n",
"\n",
"mpl.figure.prototype.handle_resize = function(fig, msg) {\n",
" var size = msg['size'];\n",
" if (size[0] != fig.canvas.width || size[1] != fig.canvas.height) {\n",
" fig._resize_canvas(size[0], size[1]);\n",
" fig.send_message(\"refresh\", {});\n",
" };\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_rubberband = function(fig, msg) {\n",
" var x0 = msg['x0'] / mpl.ratio;\n",
" var y0 = (fig.canvas.height - msg['y0']) / mpl.ratio;\n",
" var x1 = msg['x1'] / mpl.ratio;\n",
" var y1 = (fig.canvas.height - msg['y1']) / mpl.ratio;\n",
" x0 = Math.floor(x0) + 0.5;\n",
" y0 = Math.floor(y0) + 0.5;\n",
" x1 = Math.floor(x1) + 0.5;\n",
" y1 = Math.floor(y1) + 0.5;\n",
" var min_x = Math.min(x0, x1);\n",
" var min_y = Math.min(y0, y1);\n",
" var width = Math.abs(x1 - x0);\n",
" var height = Math.abs(y1 - y0);\n",
"\n",
" fig.rubberband_context.clearRect(\n",
" 0, 0, fig.canvas.width, fig.canvas.height);\n",
"\n",
" fig.rubberband_context.strokeRect(min_x, min_y, width, height);\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_figure_label = function(fig, msg) {\n",
" // Updates the figure title.\n",
" fig.header.textContent = msg['label'];\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_cursor = function(fig, msg) {\n",
" var cursor = msg['cursor'];\n",
" switch(cursor)\n",
" {\n",
" case 0:\n",
" cursor = 'pointer';\n",
" break;\n",
" case 1:\n",
" cursor = 'default';\n",
" break;\n",
" case 2:\n",
" cursor = 'crosshair';\n",
" break;\n",
" case 3:\n",
" cursor = 'move';\n",
" break;\n",
" }\n",
" fig.rubberband_canvas.style.cursor = cursor;\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_message = function(fig, msg) {\n",
" fig.message.textContent = msg['message'];\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_draw = function(fig, msg) {\n",
" // Request the server to send over a new figure.\n",
" fig.send_draw_message();\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_image_mode = function(fig, msg) {\n",
" fig.image_mode = msg['mode'];\n",
"}\n",
"\n",
"mpl.figure.prototype.updated_canvas_event = function() {\n",
" // Called whenever the canvas gets updated.\n",
" this.send_message(\"ack\", {});\n",
"}\n",
"\n",
"// A function to construct a web socket function for onmessage handling.\n",
"// Called in the figure constructor.\n",
"mpl.figure.prototype._make_on_message_function = function(fig) {\n",
" return function socket_on_message(evt) {\n",
" if (evt.data instanceof Blob) {\n",
" /* FIXME: We get \"Resource interpreted as Image but\n",
" * transferred with MIME type text/plain:\" errors on\n",
" * Chrome. But how to set the MIME type? It doesn't seem\n",
" * to be part of the websocket stream */\n",
" evt.data.type = \"image/png\";\n",
"\n",
" /* Free the memory for the previous frames */\n",
" if (fig.imageObj.src) {\n",
" (window.URL || window.webkitURL).revokeObjectURL(\n",
" fig.imageObj.src);\n",
" }\n",
"\n",
" fig.imageObj.src = (window.URL || window.webkitURL).createObjectURL(\n",
" evt.data);\n",
" fig.updated_canvas_event();\n",
" fig.waiting = false;\n",
" return;\n",
" }\n",
" else if (typeof evt.data === 'string' && evt.data.slice(0, 21) == \"data:image/png;base64\") {\n",
" fig.imageObj.src = evt.data;\n",
" fig.updated_canvas_event();\n",
" fig.waiting = false;\n",
" return;\n",
" }\n",
"\n",
" var msg = JSON.parse(evt.data);\n",
" var msg_type = msg['type'];\n",
"\n",
" // Call the \"handle_{type}\" callback, which takes\n",
" // the figure and JSON message as its only arguments.\n",
" try {\n",
" var callback = fig[\"handle_\" + msg_type];\n",
" } catch (e) {\n",
" console.log(\"No handler for the '\" + msg_type + \"' message type: \", msg);\n",
" return;\n",
" }\n",
"\n",
" if (callback) {\n",
" try {\n",
" // console.log(\"Handling '\" + msg_type + \"' message: \", msg);\n",
" callback(fig, msg);\n",
" } catch (e) {\n",
" console.log(\"Exception inside the 'handler_\" + msg_type + \"' callback:\", e, e.stack, msg);\n",
" }\n",
" }\n",
" };\n",
"}\n",
"\n",
"// from http://stackoverflow.com/questions/1114465/getting-mouse-location-in-canvas\n",
"mpl.findpos = function(e) {\n",
" //this section is from http://www.quirksmode.org/js/events_properties.html\n",
" var targ;\n",
" if (!e)\n",
" e = window.event;\n",
" if (e.target)\n",
" targ = e.target;\n",
" else if (e.srcElement)\n",
" targ = e.srcElement;\n",
" if (targ.nodeType == 3) // defeat Safari bug\n",
" targ = targ.parentNode;\n",
"\n",
" // jQuery normalizes the pageX and pageY\n",
" // pageX,Y are the mouse positions relative to the document\n",
" // offset() returns the position of the element relative to the document\n",
" var x = e.pageX - $(targ).offset().left;\n",
" var y = e.pageY - $(targ).offset().top;\n",
"\n",
" return {\"x\": x, \"y\": y};\n",
"};\n",
"\n",
"/*\n",
" * return a copy of an object with only non-object keys\n",
" * we need this to avoid circular references\n",
" * http://stackoverflow.com/a/24161582/3208463\n",
" */\n",
"function simpleKeys (original) {\n",
" return Object.keys(original).reduce(function (obj, key) {\n",
" if (typeof original[key] !== 'object')\n",
" obj[key] = original[key]\n",
" return obj;\n",
" }, {});\n",
"}\n",
"\n",
"mpl.figure.prototype.mouse_event = function(event, name) {\n",
" var canvas_pos = mpl.findpos(event)\n",
"\n",
" if (name === 'button_press')\n",
" {\n",
" this.canvas.focus();\n",
" this.canvas_div.focus();\n",
" }\n",
"\n",
" var x = canvas_pos.x * mpl.ratio;\n",
" var y = canvas_pos.y * mpl.ratio;\n",
"\n",
" this.send_message(name, {x: x, y: y, button: event.button,\n",
" step: event.step,\n",
" guiEvent: simpleKeys(event)});\n",
"\n",
" /* This prevents the web browser from automatically changing to\n",
" * the text insertion cursor when the button is pressed. We want\n",
" * to control all of the cursor setting manually through the\n",
" * 'cursor' event from matplotlib */\n",
" event.preventDefault();\n",
" return false;\n",
"}\n",
"\n",
"mpl.figure.prototype._key_event_extra = function(event, name) {\n",
" // Handle any extra behaviour associated with a key event\n",
"}\n",
"\n",
"mpl.figure.prototype.key_event = function(event, name) {\n",
"\n",
" // Prevent repeat events\n",
" if (name == 'key_press')\n",
" {\n",
" if (event.which === this._key)\n",
" return;\n",
" else\n",
" this._key = event.which;\n",
" }\n",
" if (name == 'key_release')\n",
" this._key = null;\n",
"\n",
" var value = '';\n",
" if (event.ctrlKey && event.which != 17)\n",
" value += \"ctrl+\";\n",
" if (event.altKey && event.which != 18)\n",
" value += \"alt+\";\n",
" if (event.shiftKey && event.which != 16)\n",
" value += \"shift+\";\n",
"\n",
" value += 'k';\n",
" value += event.which.toString();\n",
"\n",
" this._key_event_extra(event, name);\n",
"\n",
" this.send_message(name, {key: value,\n",
" guiEvent: simpleKeys(event)});\n",
" return false;\n",
"}\n",
"\n",
"mpl.figure.prototype.toolbar_button_onclick = function(name) {\n",
" if (name == 'download') {\n",
" this.handle_save(this, null);\n",
" } else {\n",
" this.send_message(\"toolbar_button\", {name: name});\n",
" }\n",
"};\n",
"\n",
"mpl.figure.prototype.toolbar_button_onmouseover = function(tooltip) {\n",
" this.message.textContent = tooltip;\n",
"};\n",
"mpl.toolbar_items = [[\"Home\", \"Reset original view\", \"fa fa-home icon-home\", \"home\"], [\"Back\", \"Back to previous view\", \"fa fa-arrow-left icon-arrow-left\", \"back\"], [\"Forward\", \"Forward to next view\", \"fa fa-arrow-right icon-arrow-right\", \"forward\"], [\"\", \"\", \"\", \"\"], [\"Pan\", \"Pan axes with left mouse, zoom with right\", \"fa fa-arrows icon-move\", \"pan\"], [\"Zoom\", \"Zoom to rectangle\", \"fa fa-square-o icon-check-empty\", \"zoom\"], [\"\", \"\", \"\", \"\"], [\"Download\", \"Download plot\", \"fa fa-floppy-o icon-save\", \"download\"]];\n",
"\n",
"mpl.extensions = [\"eps\", \"jpeg\", \"pdf\", \"png\", \"ps\", \"raw\", \"svg\", \"tif\"];\n",
"\n",
"mpl.default_extension = \"png\";var comm_websocket_adapter = function(comm) {\n",
" // Create a \"websocket\"-like object which calls the given IPython comm\n",
" // object with the appropriate methods. Currently this is a non binary\n",
" // socket, so there is still some room for performance tuning.\n",
" var ws = {};\n",
"\n",
" ws.close = function() {\n",
" comm.close()\n",
" };\n",
" ws.send = function(m) {\n",
" //console.log('sending', m);\n",
" comm.send(m);\n",
" };\n",
" // Register the callback with on_msg.\n",
" comm.on_msg(function(msg) {\n",
" //console.log('receiving', msg['content']['data'], msg);\n",
" // Pass the mpl event to the overridden (by mpl) onmessage function.\n",
" ws.onmessage(msg['content']['data'])\n",
" });\n",
" return ws;\n",
"}\n",
"\n",
"mpl.mpl_figure_comm = function(comm, msg) {\n",
" // This is the function which gets called when the mpl process\n",
" // starts-up an IPython Comm through the \"matplotlib\" channel.\n",
"\n",
" var id = msg.content.data.id;\n",
" // Get hold of the div created by the display call when the Comm\n",
" // socket was opened in Python.\n",
" var element = $(\"#\" + id);\n",
" var ws_proxy = comm_websocket_adapter(comm)\n",
"\n",
" function ondownload(figure, format) {\n",
" window.open(figure.imageObj.src);\n",
" }\n",
"\n",
" var fig = new mpl.figure(id, ws_proxy,\n",
" ondownload,\n",
" element.get(0));\n",
"\n",
" // Call onopen now - mpl needs it, as it is assuming we've passed it a real\n",
" // web socket which is closed, not our websocket->open comm proxy.\n",
" ws_proxy.onopen();\n",
"\n",
" fig.parent_element = element.get(0);\n",
" fig.cell_info = mpl.find_output_cell(\"\");\n",
" if (!fig.cell_info) {\n",
" console.error(\"Failed to find cell for figure\", id, fig);\n",
" return;\n",
" }\n",
"\n",
" var output_index = fig.cell_info[2]\n",
" var cell = fig.cell_info[0];\n",
"\n",
"};\n",
"\n",
"mpl.figure.prototype.handle_close = function(fig, msg) {\n",
" var width = fig.canvas.width/mpl.ratio\n",
" fig.root.unbind('remove')\n",
"\n",
" // Update the output cell to use the data from the current canvas.\n",
" fig.push_to_output();\n",
" var dataURL = fig.canvas.toDataURL();\n",
" // Re-enable the keyboard manager in IPython - without this line, in FF,\n",
" // the notebook keyboard shortcuts fail.\n",
" IPython.keyboard_manager.enable()\n",
" $(fig.parent_element).html('![](\"')
');\n",
" fig.close_ws(fig, msg);\n",
"}\n",
"\n",
"mpl.figure.prototype.close_ws = function(fig, msg){\n",
" fig.send_message('closing', msg);\n",
" // fig.ws.close()\n",
"}\n",
"\n",
"mpl.figure.prototype.push_to_output = function(remove_interactive) {\n",
" // Turn the data on the canvas into data in the output cell.\n",
" var width = this.canvas.width/mpl.ratio\n",
" var dataURL = this.canvas.toDataURL();\n",
" this.cell_info[1]['text/html'] = '';\n",
"}\n",
"\n",
"mpl.figure.prototype.updated_canvas_event = function() {\n",
" // Tell IPython that the notebook contents must change.\n",
" IPython.notebook.set_dirty(true);\n",
" this.send_message(\"ack\", {});\n",
" var fig = this;\n",
" // Wait a second, then push the new image to the DOM so\n",
" // that it is saved nicely (might be nice to debounce this).\n",
" setTimeout(function () { fig.push_to_output() }, 1000);\n",
"}\n",
"\n",
"mpl.figure.prototype._init_toolbar = function() {\n",
" var fig = this;\n",
"\n",
" var nav_element = $('')\n",
" nav_element.attr('style', 'width: 100%');\n",
" this.root.append(nav_element);\n",
"\n",
" // Define a callback function for later on.\n",
" function toolbar_event(event) {\n",
" return fig.toolbar_button_onclick(event['data']);\n",
" }\n",
" function toolbar_mouse_event(event) {\n",
" return fig.toolbar_button_onmouseover(event['data']);\n",
" }\n",
"\n",
" for(var toolbar_ind in mpl.toolbar_items){\n",
" var name = mpl.toolbar_items[toolbar_ind][0];\n",
" var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
" var image = mpl.toolbar_items[toolbar_ind][2];\n",
" var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
"\n",
" if (!name) { continue; };\n",
"\n",
" var button = $('');\n",
" button.click(method_name, toolbar_event);\n",
" button.mouseover(tooltip, toolbar_mouse_event);\n",
" nav_element.append(button);\n",
" }\n",
"\n",
" // Add the status bar.\n",
" var status_bar = $('');\n",
" nav_element.append(status_bar);\n",
" this.message = status_bar[0];\n",
"\n",
" // Add the close button to the window.\n",
" var buttongrp = $('');\n",
" var button = $('');\n",
" button.click(function (evt) { fig.handle_close(fig, {}); } );\n",
" button.mouseover('Stop Interaction', toolbar_mouse_event);\n",
" buttongrp.append(button);\n",
" var titlebar = this.root.find($('.ui-dialog-titlebar'));\n",
" titlebar.prepend(buttongrp);\n",
"}\n",
"\n",
"mpl.figure.prototype._root_extra_style = function(el){\n",
" var fig = this\n",
" el.on(\"remove\", function(){\n",
"\tfig.close_ws(fig, {});\n",
" });\n",
"}\n",
"\n",
"mpl.figure.prototype._canvas_extra_style = function(el){\n",
" // this is important to make the div 'focusable\n",
" el.attr('tabindex', 0)\n",
" // reach out to IPython and tell the keyboard manager to turn it's self\n",
" // off when our div gets focus\n",
"\n",
" // location in version 3\n",
" if (IPython.notebook.keyboard_manager) {\n",
" IPython.notebook.keyboard_manager.register_events(el);\n",
" }\n",
" else {\n",
" // location in version 2\n",
" IPython.keyboard_manager.register_events(el);\n",
" }\n",
"\n",
"}\n",
"\n",
"mpl.figure.prototype._key_event_extra = function(event, name) {\n",
" var manager = IPython.notebook.keyboard_manager;\n",
" if (!manager)\n",
" manager = IPython.keyboard_manager;\n",
"\n",
" // Check for shift+enter\n",
" if (event.shiftKey && event.which == 13) {\n",
" this.canvas_div.blur();\n",
" event.shiftKey = false;\n",
" // Send a \"J\" for go to next cell\n",
" event.which = 74;\n",
" event.keyCode = 74;\n",
" manager.command_mode();\n",
" manager.handle_keydown(event);\n",
" }\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_save = function(fig, msg) {\n",
" fig.ondownload(fig, null);\n",
"}\n",
"\n",
"\n",
"mpl.find_output_cell = function(html_output) {\n",
" // Return the cell and output element which can be found *uniquely* in the notebook.\n",
" // Note - this is a bit hacky, but it is done because the \"notebook_saving.Notebook\"\n",
" // IPython event is triggered only after the cells have been serialised, which for\n",
" // our purposes (turning an active figure into a static one), is too late.\n",
" var cells = IPython.notebook.get_cells();\n",
" var ncells = cells.length;\n",
" for (var i=0; i= 3 moved mimebundle to data attribute of output\n",
" data = data.data;\n",
" }\n",
" if (data['text/html'] == html_output) {\n",
" return [cell, data, j];\n",
" }\n",
" }\n",
" }\n",
" }\n",
"}\n",
"\n",
"// Register the function which deals with the matplotlib target/channel.\n",
"// The kernel may be null if the page has been refreshed.\n",
"if (IPython.notebook.kernel != null) {\n",
" IPython.notebook.kernel.comm_manager.register_target('matplotlib', mpl.mpl_figure_comm);\n",
"}\n"
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/html": [
""
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"# Load in our image and create a noisy version\n",
"im = np.array(Image.open('trinity.jpg')).astype('float')\n",
"np.random.seed(100)\n",
"im_noise = im + 20 * np.random.randn(*im.shape)\n",
"wave = 'db1'\n",
"\n",
"fig, ax = plt.subplots(1,2, figsize=(8,4))\n",
"ax[0].imshow(plotters.normalize(im))\n",
"ax[1].imshow(plotters.normalize(im_noise))\n",
"plt.tight_layout()"
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {
"ExecuteTime": {
"end_time": "2018-12-03T16:51:22.052399Z",
"start_time": "2018-12-03T16:51:22.004257Z"
}
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"SNR on our noisy image is: 12.673146729916327\n"
]
}
],
"source": [
"# Define our signal to noise ratio function\n",
"def snr(x, y):\n",
" ϵ = y - x\n",
" Ex = np.mean(ϵ)\n",
" Ex2 = np.mean(ϵ**2)\n",
" std = np.sqrt(Ex2 - Ex**2)\n",
" σ = np.std(x)\n",
" \n",
" snr = 20*np.log10(σ/std)\n",
" return snr\n",
"\n",
"print('SNR on our noisy image is: {}'.format(snr(im, im_noise)))"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# Use Chang's Threshold to Denoise\n",
"From equation 12 in [Adaptive Wavelet Thresholding for Image Denoising and Compression](https://ieeexplore.ieee.org/document/862633), the wavelet soft thresholds that minimise the Bayes risk for the assumption that noise is gaussian are:\n",
"$$\n",
"T_B(\\sigma_x) = \\frac{\\sigma^2}{\\sigma_x}\n",
"$$\n",
"\n",
"Where the $B$ subscript indicates that this is a function of subband, $\\sigma_x$ is the variance of the noiseless wavelet coefficients, and $\\sigma$ is the variance of the noise.\n",
"\n",
"Note that we can estimate $\\sigma$ robustly by calculating the median absolute deviation of the high-high wavelet coefficients at the finest level.\n",
"\n",
"$$\n",
"\\sigma \\approx \\frac{\\text{median}(|x|)}{.6745}\n",
"$$\n",
"\n",
"Then\n",
"\n",
"$$\n",
"\\sigma_X = \\sqrt{\\sigma_Y^2 - \\sigma^2}\n",
"$$"
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {
"ExecuteTime": {
"end_time": "2018-12-03T16:51:22.100992Z",
"start_time": "2018-12-03T16:51:22.053619Z"
}
},
"outputs": [],
"source": [
"def bayes_thresh(im, J=3):\n",
" \"\"\" Calculates the soft shrink thresholds for the 3 subbands of a J\n",
" level transform. the input image can have more than one channel. The\n",
" resulting array has shape\n",
" \n",
" T[J, 3, C], where the zeroth dimension iterates of the scales, the first \n",
" dimension iterates over the bands LH, HL and HH, and the final dimension \n",
" iterates over the channels.\"\"\"\n",
" C = im.shape[-1]\n",
" coeffs = pywt.wavedec2(im, wave, axes=(0, 1), level=J)\n",
"\n",
" # Estimate the noise variance using the median\n",
" σ = np.zeros((C))\n",
" for c in range(C):\n",
" σ[c] = np.median(np.abs(coeffs[-1][-1][:, :, c].ravel())) / .6745\n",
"\n",
" # Estimate the noisy signal variance\n",
" σy2 = np.zeros((J, 3, C))\n",
" for j in range(J):\n",
" for b, band in enumerate(['LH', 'HL', 'HH']):\n",
" for c in range(C):\n",
" σy2[j, b, c] = np.sum(coeffs[1 + j][b][:, :, c]**2) / \\\n",
" coeffs[1 + j][b][:, :, c].size\n",
"\n",
" # Calculate σ_x = sqrt(σ_y^2 - σ^2)\n",
" σx = np.sqrt(np.maximum(σy2 - σ**2, 0.0001))\n",
"\n",
" # Calculate T\n",
" T = σ**2 / σx\n",
" return T\n",
"\n",
"def shrink(x, t):\n",
" \"\"\" Given a wavelet coefficient and a threshold, shrink \"\"\"\n",
" if t == 0:\n",
" return x\n",
" m = np.abs(x)\n",
" denom = m + (m < t).astype('float')\n",
" gain = np.maximum(m - t, 0)/denom\n",
" return x * gain\n",
" \n",
"def shrink_coeffs(coeffs, T):\n",
" \"\"\" Shrink the wavelet coefficients with the thresholds T. T should be an \n",
" array of shape (J, 3, C) for color images.\"\"\"\n",
" assert T.shape[0] == len(coeffs) - 1\n",
" J = len(coeffs) - 1\n",
" assert T.shape[1] == len(coeffs[1])\n",
" assert T.shape[2] == len(coeffs[1][0][0,0])\n",
" C = T.shape[2]\n",
" \n",
" coeffs_new = [None,] * (J + 1)\n",
" coeffs_new[0] = np.copy(coeffs[0])\n",
" for j in range(J):\n",
" coeffs_new[1+j] = [np.zeros_like(coeffs[1+j][0]),\n",
" np.zeros_like(coeffs[1+j][1]),\n",
" np.zeros_like(coeffs[1+j][2])] \n",
" for b, band in enumerate(['LH', 'HL', 'HH']):\n",
" for c in range(C):\n",
" coeffs_new[1+j][b][:,:,c] = shrink(coeffs[1+j][b][:,:,c], T[j,b,c])\n",
" \n",
" return coeffs_new\n"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Use the above to denoise the image"
]
},
{
"cell_type": "code",
"execution_count": 5,
"metadata": {
"ExecuteTime": {
"end_time": "2018-12-03T16:51:22.917813Z",
"start_time": "2018-12-03T16:51:22.102355Z"
}
},
"outputs": [
{
"data": {
"application/javascript": [
"/* Put everything inside the global mpl namespace */\n",
"window.mpl = {};\n",
"\n",
"\n",
"mpl.get_websocket_type = function() {\n",
" if (typeof(WebSocket) !== 'undefined') {\n",
" return WebSocket;\n",
" } else if (typeof(MozWebSocket) !== 'undefined') {\n",
" return MozWebSocket;\n",
" } else {\n",
" alert('Your browser does not have WebSocket support.' +\n",
" 'Please try Chrome, Safari or Firefox ≥ 6. ' +\n",
" 'Firefox 4 and 5 are also supported but you ' +\n",
" 'have to enable WebSockets in about:config.');\n",
" };\n",
"}\n",
"\n",
"mpl.figure = function(figure_id, websocket, ondownload, parent_element) {\n",
" this.id = figure_id;\n",
"\n",
" this.ws = websocket;\n",
"\n",
" this.supports_binary = (this.ws.binaryType != undefined);\n",
"\n",
" if (!this.supports_binary) {\n",
" var warnings = document.getElementById(\"mpl-warnings\");\n",
" if (warnings) {\n",
" warnings.style.display = 'block';\n",
" warnings.textContent = (\n",
" \"This browser does not support binary websocket messages. \" +\n",
" \"Performance may be slow.\");\n",
" }\n",
" }\n",
"\n",
" this.imageObj = new Image();\n",
"\n",
" this.context = undefined;\n",
" this.message = undefined;\n",
" this.canvas = undefined;\n",
" this.rubberband_canvas = undefined;\n",
" this.rubberband_context = undefined;\n",
" this.format_dropdown = undefined;\n",
"\n",
" this.image_mode = 'full';\n",
"\n",
" this.root = $('');\n",
" this._root_extra_style(this.root)\n",
" this.root.attr('style', 'display: inline-block');\n",
"\n",
" $(parent_element).append(this.root);\n",
"\n",
" this._init_header(this);\n",
" this._init_canvas(this);\n",
" this._init_toolbar(this);\n",
"\n",
" var fig = this;\n",
"\n",
" this.waiting = false;\n",
"\n",
" this.ws.onopen = function () {\n",
" fig.send_message(\"supports_binary\", {value: fig.supports_binary});\n",
" fig.send_message(\"send_image_mode\", {});\n",
" if (mpl.ratio != 1) {\n",
" fig.send_message(\"set_dpi_ratio\", {'dpi_ratio': mpl.ratio});\n",
" }\n",
" fig.send_message(\"refresh\", {});\n",
" }\n",
"\n",
" this.imageObj.onload = function() {\n",
" if (fig.image_mode == 'full') {\n",
" // Full images could contain transparency (where diff images\n",
" // almost always do), so we need to clear the canvas so that\n",
" // there is no ghosting.\n",
" fig.context.clearRect(0, 0, fig.canvas.width, fig.canvas.height);\n",
" }\n",
" fig.context.drawImage(fig.imageObj, 0, 0);\n",
" };\n",
"\n",
" this.imageObj.onunload = function() {\n",
" fig.ws.close();\n",
" }\n",
"\n",
" this.ws.onmessage = this._make_on_message_function(this);\n",
"\n",
" this.ondownload = ondownload;\n",
"}\n",
"\n",
"mpl.figure.prototype._init_header = function() {\n",
" var titlebar = $(\n",
" '');\n",
" var titletext = $(\n",
" '');\n",
" titlebar.append(titletext)\n",
" this.root.append(titlebar);\n",
" this.header = titletext[0];\n",
"}\n",
"\n",
"\n",
"\n",
"mpl.figure.prototype._canvas_extra_style = function(canvas_div) {\n",
"\n",
"}\n",
"\n",
"\n",
"mpl.figure.prototype._root_extra_style = function(canvas_div) {\n",
"\n",
"}\n",
"\n",
"mpl.figure.prototype._init_canvas = function() {\n",
" var fig = this;\n",
"\n",
" var canvas_div = $('');\n",
"\n",
" canvas_div.attr('style', 'position: relative; clear: both; outline: 0');\n",
"\n",
" function canvas_keyboard_event(event) {\n",
" return fig.key_event(event, event['data']);\n",
" }\n",
"\n",
" canvas_div.keydown('key_press', canvas_keyboard_event);\n",
" canvas_div.keyup('key_release', canvas_keyboard_event);\n",
" this.canvas_div = canvas_div\n",
" this._canvas_extra_style(canvas_div)\n",
" this.root.append(canvas_div);\n",
"\n",
" var canvas = $('');\n",
" canvas.addClass('mpl-canvas');\n",
" canvas.attr('style', \"left: 0; top: 0; z-index: 0; outline: 0\")\n",
"\n",
" this.canvas = canvas[0];\n",
" this.context = canvas[0].getContext(\"2d\");\n",
"\n",
" var backingStore = this.context.backingStorePixelRatio ||\n",
"\tthis.context.webkitBackingStorePixelRatio ||\n",
"\tthis.context.mozBackingStorePixelRatio ||\n",
"\tthis.context.msBackingStorePixelRatio ||\n",
"\tthis.context.oBackingStorePixelRatio ||\n",
"\tthis.context.backingStorePixelRatio || 1;\n",
"\n",
" mpl.ratio = (window.devicePixelRatio || 1) / backingStore;\n",
"\n",
" var rubberband = $('');\n",
" rubberband.attr('style', \"position: absolute; left: 0; top: 0; z-index: 1;\")\n",
"\n",
" var pass_mouse_events = true;\n",
"\n",
" canvas_div.resizable({\n",
" start: function(event, ui) {\n",
" pass_mouse_events = false;\n",
" },\n",
" resize: function(event, ui) {\n",
" fig.request_resize(ui.size.width, ui.size.height);\n",
" },\n",
" stop: function(event, ui) {\n",
" pass_mouse_events = true;\n",
" fig.request_resize(ui.size.width, ui.size.height);\n",
" },\n",
" });\n",
"\n",
" function mouse_event_fn(event) {\n",
" if (pass_mouse_events)\n",
" return fig.mouse_event(event, event['data']);\n",
" }\n",
"\n",
" rubberband.mousedown('button_press', mouse_event_fn);\n",
" rubberband.mouseup('button_release', mouse_event_fn);\n",
" // Throttle sequential mouse events to 1 every 20ms.\n",
" rubberband.mousemove('motion_notify', mouse_event_fn);\n",
"\n",
" rubberband.mouseenter('figure_enter', mouse_event_fn);\n",
" rubberband.mouseleave('figure_leave', mouse_event_fn);\n",
"\n",
" canvas_div.on(\"wheel\", function (event) {\n",
" event = event.originalEvent;\n",
" event['data'] = 'scroll'\n",
" if (event.deltaY < 0) {\n",
" event.step = 1;\n",
" } else {\n",
" event.step = -1;\n",
" }\n",
" mouse_event_fn(event);\n",
" });\n",
"\n",
" canvas_div.append(canvas);\n",
" canvas_div.append(rubberband);\n",
"\n",
" this.rubberband = rubberband;\n",
" this.rubberband_canvas = rubberband[0];\n",
" this.rubberband_context = rubberband[0].getContext(\"2d\");\n",
" this.rubberband_context.strokeStyle = \"#000000\";\n",
"\n",
" this._resize_canvas = function(width, height) {\n",
" // Keep the size of the canvas, canvas container, and rubber band\n",
" // canvas in synch.\n",
" canvas_div.css('width', width)\n",
" canvas_div.css('height', height)\n",
"\n",
" canvas.attr('width', width * mpl.ratio);\n",
" canvas.attr('height', height * mpl.ratio);\n",
" canvas.attr('style', 'width: ' + width + 'px; height: ' + height + 'px;');\n",
"\n",
" rubberband.attr('width', width);\n",
" rubberband.attr('height', height);\n",
" }\n",
"\n",
" // Set the figure to an initial 600x600px, this will subsequently be updated\n",
" // upon first draw.\n",
" this._resize_canvas(600, 600);\n",
"\n",
" // Disable right mouse context menu.\n",
" $(this.rubberband_canvas).bind(\"contextmenu\",function(e){\n",
" return false;\n",
" });\n",
"\n",
" function set_focus () {\n",
" canvas.focus();\n",
" canvas_div.focus();\n",
" }\n",
"\n",
" window.setTimeout(set_focus, 100);\n",
"}\n",
"\n",
"mpl.figure.prototype._init_toolbar = function() {\n",
" var fig = this;\n",
"\n",
" var nav_element = $('')\n",
" nav_element.attr('style', 'width: 100%');\n",
" this.root.append(nav_element);\n",
"\n",
" // Define a callback function for later on.\n",
" function toolbar_event(event) {\n",
" return fig.toolbar_button_onclick(event['data']);\n",
" }\n",
" function toolbar_mouse_event(event) {\n",
" return fig.toolbar_button_onmouseover(event['data']);\n",
" }\n",
"\n",
" for(var toolbar_ind in mpl.toolbar_items) {\n",
" var name = mpl.toolbar_items[toolbar_ind][0];\n",
" var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
" var image = mpl.toolbar_items[toolbar_ind][2];\n",
" var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
"\n",
" if (!name) {\n",
" // put a spacer in here.\n",
" continue;\n",
" }\n",
" var button = $('');\n",
" button.addClass('ui-button ui-widget ui-state-default ui-corner-all ' +\n",
" 'ui-button-icon-only');\n",
" button.attr('role', 'button');\n",
" button.attr('aria-disabled', 'false');\n",
" button.click(method_name, toolbar_event);\n",
" button.mouseover(tooltip, toolbar_mouse_event);\n",
"\n",
" var icon_img = $('');\n",
" icon_img.addClass('ui-button-icon-primary ui-icon');\n",
" icon_img.addClass(image);\n",
" icon_img.addClass('ui-corner-all');\n",
"\n",
" var tooltip_span = $('');\n",
" tooltip_span.addClass('ui-button-text');\n",
" tooltip_span.html(tooltip);\n",
"\n",
" button.append(icon_img);\n",
" button.append(tooltip_span);\n",
"\n",
" nav_element.append(button);\n",
" }\n",
"\n",
" var fmt_picker_span = $('');\n",
"\n",
" var fmt_picker = $('');\n",
" fmt_picker.addClass('mpl-toolbar-option ui-widget ui-widget-content');\n",
" fmt_picker_span.append(fmt_picker);\n",
" nav_element.append(fmt_picker_span);\n",
" this.format_dropdown = fmt_picker[0];\n",
"\n",
" for (var ind in mpl.extensions) {\n",
" var fmt = mpl.extensions[ind];\n",
" var option = $(\n",
" '', {selected: fmt === mpl.default_extension}).html(fmt);\n",
" fmt_picker.append(option)\n",
" }\n",
"\n",
" // Add hover states to the ui-buttons\n",
" $( \".ui-button\" ).hover(\n",
" function() { $(this).addClass(\"ui-state-hover\");},\n",
" function() { $(this).removeClass(\"ui-state-hover\");}\n",
" );\n",
"\n",
" var status_bar = $('');\n",
" nav_element.append(status_bar);\n",
" this.message = status_bar[0];\n",
"}\n",
"\n",
"mpl.figure.prototype.request_resize = function(x_pixels, y_pixels) {\n",
" // Request matplotlib to resize the figure. Matplotlib will then trigger a resize in the client,\n",
" // which will in turn request a refresh of the image.\n",
" this.send_message('resize', {'width': x_pixels, 'height': y_pixels});\n",
"}\n",
"\n",
"mpl.figure.prototype.send_message = function(type, properties) {\n",
" properties['type'] = type;\n",
" properties['figure_id'] = this.id;\n",
" this.ws.send(JSON.stringify(properties));\n",
"}\n",
"\n",
"mpl.figure.prototype.send_draw_message = function() {\n",
" if (!this.waiting) {\n",
" this.waiting = true;\n",
" this.ws.send(JSON.stringify({type: \"draw\", figure_id: this.id}));\n",
" }\n",
"}\n",
"\n",
"\n",
"mpl.figure.prototype.handle_save = function(fig, msg) {\n",
" var format_dropdown = fig.format_dropdown;\n",
" var format = format_dropdown.options[format_dropdown.selectedIndex].value;\n",
" fig.ondownload(fig, format);\n",
"}\n",
"\n",
"\n",
"mpl.figure.prototype.handle_resize = function(fig, msg) {\n",
" var size = msg['size'];\n",
" if (size[0] != fig.canvas.width || size[1] != fig.canvas.height) {\n",
" fig._resize_canvas(size[0], size[1]);\n",
" fig.send_message(\"refresh\", {});\n",
" };\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_rubberband = function(fig, msg) {\n",
" var x0 = msg['x0'] / mpl.ratio;\n",
" var y0 = (fig.canvas.height - msg['y0']) / mpl.ratio;\n",
" var x1 = msg['x1'] / mpl.ratio;\n",
" var y1 = (fig.canvas.height - msg['y1']) / mpl.ratio;\n",
" x0 = Math.floor(x0) + 0.5;\n",
" y0 = Math.floor(y0) + 0.5;\n",
" x1 = Math.floor(x1) + 0.5;\n",
" y1 = Math.floor(y1) + 0.5;\n",
" var min_x = Math.min(x0, x1);\n",
" var min_y = Math.min(y0, y1);\n",
" var width = Math.abs(x1 - x0);\n",
" var height = Math.abs(y1 - y0);\n",
"\n",
" fig.rubberband_context.clearRect(\n",
" 0, 0, fig.canvas.width, fig.canvas.height);\n",
"\n",
" fig.rubberband_context.strokeRect(min_x, min_y, width, height);\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_figure_label = function(fig, msg) {\n",
" // Updates the figure title.\n",
" fig.header.textContent = msg['label'];\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_cursor = function(fig, msg) {\n",
" var cursor = msg['cursor'];\n",
" switch(cursor)\n",
" {\n",
" case 0:\n",
" cursor = 'pointer';\n",
" break;\n",
" case 1:\n",
" cursor = 'default';\n",
" break;\n",
" case 2:\n",
" cursor = 'crosshair';\n",
" break;\n",
" case 3:\n",
" cursor = 'move';\n",
" break;\n",
" }\n",
" fig.rubberband_canvas.style.cursor = cursor;\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_message = function(fig, msg) {\n",
" fig.message.textContent = msg['message'];\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_draw = function(fig, msg) {\n",
" // Request the server to send over a new figure.\n",
" fig.send_draw_message();\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_image_mode = function(fig, msg) {\n",
" fig.image_mode = msg['mode'];\n",
"}\n",
"\n",
"mpl.figure.prototype.updated_canvas_event = function() {\n",
" // Called whenever the canvas gets updated.\n",
" this.send_message(\"ack\", {});\n",
"}\n",
"\n",
"// A function to construct a web socket function for onmessage handling.\n",
"// Called in the figure constructor.\n",
"mpl.figure.prototype._make_on_message_function = function(fig) {\n",
" return function socket_on_message(evt) {\n",
" if (evt.data instanceof Blob) {\n",
" /* FIXME: We get \"Resource interpreted as Image but\n",
" * transferred with MIME type text/plain:\" errors on\n",
" * Chrome. But how to set the MIME type? It doesn't seem\n",
" * to be part of the websocket stream */\n",
" evt.data.type = \"image/png\";\n",
"\n",
" /* Free the memory for the previous frames */\n",
" if (fig.imageObj.src) {\n",
" (window.URL || window.webkitURL).revokeObjectURL(\n",
" fig.imageObj.src);\n",
" }\n",
"\n",
" fig.imageObj.src = (window.URL || window.webkitURL).createObjectURL(\n",
" evt.data);\n",
" fig.updated_canvas_event();\n",
" fig.waiting = false;\n",
" return;\n",
" }\n",
" else if (typeof evt.data === 'string' && evt.data.slice(0, 21) == \"data:image/png;base64\") {\n",
" fig.imageObj.src = evt.data;\n",
" fig.updated_canvas_event();\n",
" fig.waiting = false;\n",
" return;\n",
" }\n",
"\n",
" var msg = JSON.parse(evt.data);\n",
" var msg_type = msg['type'];\n",
"\n",
" // Call the \"handle_{type}\" callback, which takes\n",
" // the figure and JSON message as its only arguments.\n",
" try {\n",
" var callback = fig[\"handle_\" + msg_type];\n",
" } catch (e) {\n",
" console.log(\"No handler for the '\" + msg_type + \"' message type: \", msg);\n",
" return;\n",
" }\n",
"\n",
" if (callback) {\n",
" try {\n",
" // console.log(\"Handling '\" + msg_type + \"' message: \", msg);\n",
" callback(fig, msg);\n",
" } catch (e) {\n",
" console.log(\"Exception inside the 'handler_\" + msg_type + \"' callback:\", e, e.stack, msg);\n",
" }\n",
" }\n",
" };\n",
"}\n",
"\n",
"// from http://stackoverflow.com/questions/1114465/getting-mouse-location-in-canvas\n",
"mpl.findpos = function(e) {\n",
" //this section is from http://www.quirksmode.org/js/events_properties.html\n",
" var targ;\n",
" if (!e)\n",
" e = window.event;\n",
" if (e.target)\n",
" targ = e.target;\n",
" else if (e.srcElement)\n",
" targ = e.srcElement;\n",
" if (targ.nodeType == 3) // defeat Safari bug\n",
" targ = targ.parentNode;\n",
"\n",
" // jQuery normalizes the pageX and pageY\n",
" // pageX,Y are the mouse positions relative to the document\n",
" // offset() returns the position of the element relative to the document\n",
" var x = e.pageX - $(targ).offset().left;\n",
" var y = e.pageY - $(targ).offset().top;\n",
"\n",
" return {\"x\": x, \"y\": y};\n",
"};\n",
"\n",
"/*\n",
" * return a copy of an object with only non-object keys\n",
" * we need this to avoid circular references\n",
" * http://stackoverflow.com/a/24161582/3208463\n",
" */\n",
"function simpleKeys (original) {\n",
" return Object.keys(original).reduce(function (obj, key) {\n",
" if (typeof original[key] !== 'object')\n",
" obj[key] = original[key]\n",
" return obj;\n",
" }, {});\n",
"}\n",
"\n",
"mpl.figure.prototype.mouse_event = function(event, name) {\n",
" var canvas_pos = mpl.findpos(event)\n",
"\n",
" if (name === 'button_press')\n",
" {\n",
" this.canvas.focus();\n",
" this.canvas_div.focus();\n",
" }\n",
"\n",
" var x = canvas_pos.x * mpl.ratio;\n",
" var y = canvas_pos.y * mpl.ratio;\n",
"\n",
" this.send_message(name, {x: x, y: y, button: event.button,\n",
" step: event.step,\n",
" guiEvent: simpleKeys(event)});\n",
"\n",
" /* This prevents the web browser from automatically changing to\n",
" * the text insertion cursor when the button is pressed. We want\n",
" * to control all of the cursor setting manually through the\n",
" * 'cursor' event from matplotlib */\n",
" event.preventDefault();\n",
" return false;\n",
"}\n",
"\n",
"mpl.figure.prototype._key_event_extra = function(event, name) {\n",
" // Handle any extra behaviour associated with a key event\n",
"}\n",
"\n",
"mpl.figure.prototype.key_event = function(event, name) {\n",
"\n",
" // Prevent repeat events\n",
" if (name == 'key_press')\n",
" {\n",
" if (event.which === this._key)\n",
" return;\n",
" else\n",
" this._key = event.which;\n",
" }\n",
" if (name == 'key_release')\n",
" this._key = null;\n",
"\n",
" var value = '';\n",
" if (event.ctrlKey && event.which != 17)\n",
" value += \"ctrl+\";\n",
" if (event.altKey && event.which != 18)\n",
" value += \"alt+\";\n",
" if (event.shiftKey && event.which != 16)\n",
" value += \"shift+\";\n",
"\n",
" value += 'k';\n",
" value += event.which.toString();\n",
"\n",
" this._key_event_extra(event, name);\n",
"\n",
" this.send_message(name, {key: value,\n",
" guiEvent: simpleKeys(event)});\n",
" return false;\n",
"}\n",
"\n",
"mpl.figure.prototype.toolbar_button_onclick = function(name) {\n",
" if (name == 'download') {\n",
" this.handle_save(this, null);\n",
" } else {\n",
" this.send_message(\"toolbar_button\", {name: name});\n",
" }\n",
"};\n",
"\n",
"mpl.figure.prototype.toolbar_button_onmouseover = function(tooltip) {\n",
" this.message.textContent = tooltip;\n",
"};\n",
"mpl.toolbar_items = [[\"Home\", \"Reset original view\", \"fa fa-home icon-home\", \"home\"], [\"Back\", \"Back to previous view\", \"fa fa-arrow-left icon-arrow-left\", \"back\"], [\"Forward\", \"Forward to next view\", \"fa fa-arrow-right icon-arrow-right\", \"forward\"], [\"\", \"\", \"\", \"\"], [\"Pan\", \"Pan axes with left mouse, zoom with right\", \"fa fa-arrows icon-move\", \"pan\"], [\"Zoom\", \"Zoom to rectangle\", \"fa fa-square-o icon-check-empty\", \"zoom\"], [\"\", \"\", \"\", \"\"], [\"Download\", \"Download plot\", \"fa fa-floppy-o icon-save\", \"download\"]];\n",
"\n",
"mpl.extensions = [\"eps\", \"jpeg\", \"pdf\", \"png\", \"ps\", \"raw\", \"svg\", \"tif\"];\n",
"\n",
"mpl.default_extension = \"png\";var comm_websocket_adapter = function(comm) {\n",
" // Create a \"websocket\"-like object which calls the given IPython comm\n",
" // object with the appropriate methods. Currently this is a non binary\n",
" // socket, so there is still some room for performance tuning.\n",
" var ws = {};\n",
"\n",
" ws.close = function() {\n",
" comm.close()\n",
" };\n",
" ws.send = function(m) {\n",
" //console.log('sending', m);\n",
" comm.send(m);\n",
" };\n",
" // Register the callback with on_msg.\n",
" comm.on_msg(function(msg) {\n",
" //console.log('receiving', msg['content']['data'], msg);\n",
" // Pass the mpl event to the overridden (by mpl) onmessage function.\n",
" ws.onmessage(msg['content']['data'])\n",
" });\n",
" return ws;\n",
"}\n",
"\n",
"mpl.mpl_figure_comm = function(comm, msg) {\n",
" // This is the function which gets called when the mpl process\n",
" // starts-up an IPython Comm through the \"matplotlib\" channel.\n",
"\n",
" var id = msg.content.data.id;\n",
" // Get hold of the div created by the display call when the Comm\n",
" // socket was opened in Python.\n",
" var element = $(\"#\" + id);\n",
" var ws_proxy = comm_websocket_adapter(comm)\n",
"\n",
" function ondownload(figure, format) {\n",
" window.open(figure.imageObj.src);\n",
" }\n",
"\n",
" var fig = new mpl.figure(id, ws_proxy,\n",
" ondownload,\n",
" element.get(0));\n",
"\n",
" // Call onopen now - mpl needs it, as it is assuming we've passed it a real\n",
" // web socket which is closed, not our websocket->open comm proxy.\n",
" ws_proxy.onopen();\n",
"\n",
" fig.parent_element = element.get(0);\n",
" fig.cell_info = mpl.find_output_cell(\"\");\n",
" if (!fig.cell_info) {\n",
" console.error(\"Failed to find cell for figure\", id, fig);\n",
" return;\n",
" }\n",
"\n",
" var output_index = fig.cell_info[2]\n",
" var cell = fig.cell_info[0];\n",
"\n",
"};\n",
"\n",
"mpl.figure.prototype.handle_close = function(fig, msg) {\n",
" var width = fig.canvas.width/mpl.ratio\n",
" fig.root.unbind('remove')\n",
"\n",
" // Update the output cell to use the data from the current canvas.\n",
" fig.push_to_output();\n",
" var dataURL = fig.canvas.toDataURL();\n",
" // Re-enable the keyboard manager in IPython - without this line, in FF,\n",
" // the notebook keyboard shortcuts fail.\n",
" IPython.keyboard_manager.enable()\n",
" $(fig.parent_element).html('');\n",
" fig.close_ws(fig, msg);\n",
"}\n",
"\n",
"mpl.figure.prototype.close_ws = function(fig, msg){\n",
" fig.send_message('closing', msg);\n",
" // fig.ws.close()\n",
"}\n",
"\n",
"mpl.figure.prototype.push_to_output = function(remove_interactive) {\n",
" // Turn the data on the canvas into data in the output cell.\n",
" var width = this.canvas.width/mpl.ratio\n",
" var dataURL = this.canvas.toDataURL();\n",
" this.cell_info[1]['text/html'] = '';\n",
"}\n",
"\n",
"mpl.figure.prototype.updated_canvas_event = function() {\n",
" // Tell IPython that the notebook contents must change.\n",
" IPython.notebook.set_dirty(true);\n",
" this.send_message(\"ack\", {});\n",
" var fig = this;\n",
" // Wait a second, then push the new image to the DOM so\n",
" // that it is saved nicely (might be nice to debounce this).\n",
" setTimeout(function () { fig.push_to_output() }, 1000);\n",
"}\n",
"\n",
"mpl.figure.prototype._init_toolbar = function() {\n",
" var fig = this;\n",
"\n",
" var nav_element = $('')\n",
" nav_element.attr('style', 'width: 100%');\n",
" this.root.append(nav_element);\n",
"\n",
" // Define a callback function for later on.\n",
" function toolbar_event(event) {\n",
" return fig.toolbar_button_onclick(event['data']);\n",
" }\n",
" function toolbar_mouse_event(event) {\n",
" return fig.toolbar_button_onmouseover(event['data']);\n",
" }\n",
"\n",
" for(var toolbar_ind in mpl.toolbar_items){\n",
" var name = mpl.toolbar_items[toolbar_ind][0];\n",
" var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
" var image = mpl.toolbar_items[toolbar_ind][2];\n",
" var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
"\n",
" if (!name) { continue; };\n",
"\n",
" var button = $('');\n",
" button.click(method_name, toolbar_event);\n",
" button.mouseover(tooltip, toolbar_mouse_event);\n",
" nav_element.append(button);\n",
" }\n",
"\n",
" // Add the status bar.\n",
" var status_bar = $('');\n",
" nav_element.append(status_bar);\n",
" this.message = status_bar[0];\n",
"\n",
" // Add the close button to the window.\n",
" var buttongrp = $('');\n",
" var button = $('');\n",
" button.click(function (evt) { fig.handle_close(fig, {}); } );\n",
" button.mouseover('Stop Interaction', toolbar_mouse_event);\n",
" buttongrp.append(button);\n",
" var titlebar = this.root.find($('.ui-dialog-titlebar'));\n",
" titlebar.prepend(buttongrp);\n",
"}\n",
"\n",
"mpl.figure.prototype._root_extra_style = function(el){\n",
" var fig = this\n",
" el.on(\"remove\", function(){\n",
"\tfig.close_ws(fig, {});\n",
" });\n",
"}\n",
"\n",
"mpl.figure.prototype._canvas_extra_style = function(el){\n",
" // this is important to make the div 'focusable\n",
" el.attr('tabindex', 0)\n",
" // reach out to IPython and tell the keyboard manager to turn it's self\n",
" // off when our div gets focus\n",
"\n",
" // location in version 3\n",
" if (IPython.notebook.keyboard_manager) {\n",
" IPython.notebook.keyboard_manager.register_events(el);\n",
" }\n",
" else {\n",
" // location in version 2\n",
" IPython.keyboard_manager.register_events(el);\n",
" }\n",
"\n",
"}\n",
"\n",
"mpl.figure.prototype._key_event_extra = function(event, name) {\n",
" var manager = IPython.notebook.keyboard_manager;\n",
" if (!manager)\n",
" manager = IPython.keyboard_manager;\n",
"\n",
" // Check for shift+enter\n",
" if (event.shiftKey && event.which == 13) {\n",
" this.canvas_div.blur();\n",
" event.shiftKey = false;\n",
" // Send a \"J\" for go to next cell\n",
" event.which = 74;\n",
" event.keyCode = 74;\n",
" manager.command_mode();\n",
" manager.handle_keydown(event);\n",
" }\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_save = function(fig, msg) {\n",
" fig.ondownload(fig, null);\n",
"}\n",
"\n",
"\n",
"mpl.find_output_cell = function(html_output) {\n",
" // Return the cell and output element which can be found *uniquely* in the notebook.\n",
" // Note - this is a bit hacky, but it is done because the \"notebook_saving.Notebook\"\n",
" // IPython event is triggered only after the cells have been serialised, which for\n",
" // our purposes (turning an active figure into a static one), is too late.\n",
" var cells = IPython.notebook.get_cells();\n",
" var ncells = cells.length;\n",
" for (var i=0; i= 3 moved mimebundle to data attribute of output\n",
" data = data.data;\n",
" }\n",
" if (data['text/html'] == html_output) {\n",
" return [cell, data, j];\n",
" }\n",
" }\n",
" }\n",
" }\n",
"}\n",
"\n",
"// Register the function which deals with the matplotlib target/channel.\n",
"// The kernel may be null if the page has been refreshed.\n",
"if (IPython.notebook.kernel != null) {\n",
" IPython.notebook.kernel.comm_manager.register_target('matplotlib', mpl.mpl_figure_comm);\n",
"}\n"
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/html": [
""
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/plain": [
""
]
},
"execution_count": 5,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"T = bayes_thresh(im_noise)\n",
"coeffs = pywt.wavedec2(im_noise, wave, axes=(0,1), level=3)\n",
"x_hat = pywt.waverec2(shrink_coeffs(coeffs, T), wave, axes=(0,1))\n",
"\n",
"# Plot the result\n",
"fig = plt.figure(figsize=(8,6))\n",
"gs = gridspec.GridSpec(2, 4, hspace=0.1, wspace=0.1, top=0.95, bottom=0.05)\n",
"ax1 = plt.subplot(gs[0,1:3], xticks=[], yticks=[], title='Input Image')\n",
"ax2 = plt.subplot(gs[1,:2], xticks=[], yticks=[], \n",
" title='Noisy Image (SNR={:.2f}dB)'.format(snr(im, im_noise)))\n",
"ax3 = plt.subplot(gs[1,2:], xticks=[], yticks=[], \n",
" title='Denoised Image (SNR={:.2f}dB)'.format(snr(im, x_hat)))\n",
"ax1.imshow(plotters.normalize(im))\n",
"ax2.imshow(plotters.normalize(im_noise))\n",
"ax3.imshow(plotters.normalize(x_hat))"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# Analysis and tests\n",
"How many elements are set to 0 across each of the subbands? Let us plot the distributions of the wavelet coefficients for three scales before and after shrinkage. For plotting purposes, we do not show the number of zeros in the shrunk coefficients as this will greatly outweigh all other values."
]
},
{
"cell_type": "code",
"execution_count": 6,
"metadata": {
"ExecuteTime": {
"end_time": "2018-12-03T16:51:23.834463Z",
"start_time": "2018-12-03T16:51:22.919054Z"
},
"scrolled": false
},
"outputs": [
{
"data": {
"application/javascript": [
"/* Put everything inside the global mpl namespace */\n",
"window.mpl = {};\n",
"\n",
"\n",
"mpl.get_websocket_type = function() {\n",
" if (typeof(WebSocket) !== 'undefined') {\n",
" return WebSocket;\n",
" } else if (typeof(MozWebSocket) !== 'undefined') {\n",
" return MozWebSocket;\n",
" } else {\n",
" alert('Your browser does not have WebSocket support.' +\n",
" 'Please try Chrome, Safari or Firefox ≥ 6. ' +\n",
" 'Firefox 4 and 5 are also supported but you ' +\n",
" 'have to enable WebSockets in about:config.');\n",
" };\n",
"}\n",
"\n",
"mpl.figure = function(figure_id, websocket, ondownload, parent_element) {\n",
" this.id = figure_id;\n",
"\n",
" this.ws = websocket;\n",
"\n",
" this.supports_binary = (this.ws.binaryType != undefined);\n",
"\n",
" if (!this.supports_binary) {\n",
" var warnings = document.getElementById(\"mpl-warnings\");\n",
" if (warnings) {\n",
" warnings.style.display = 'block';\n",
" warnings.textContent = (\n",
" \"This browser does not support binary websocket messages. \" +\n",
" \"Performance may be slow.\");\n",
" }\n",
" }\n",
"\n",
" this.imageObj = new Image();\n",
"\n",
" this.context = undefined;\n",
" this.message = undefined;\n",
" this.canvas = undefined;\n",
" this.rubberband_canvas = undefined;\n",
" this.rubberband_context = undefined;\n",
" this.format_dropdown = undefined;\n",
"\n",
" this.image_mode = 'full';\n",
"\n",
" this.root = $('');\n",
" this._root_extra_style(this.root)\n",
" this.root.attr('style', 'display: inline-block');\n",
"\n",
" $(parent_element).append(this.root);\n",
"\n",
" this._init_header(this);\n",
" this._init_canvas(this);\n",
" this._init_toolbar(this);\n",
"\n",
" var fig = this;\n",
"\n",
" this.waiting = false;\n",
"\n",
" this.ws.onopen = function () {\n",
" fig.send_message(\"supports_binary\", {value: fig.supports_binary});\n",
" fig.send_message(\"send_image_mode\", {});\n",
" if (mpl.ratio != 1) {\n",
" fig.send_message(\"set_dpi_ratio\", {'dpi_ratio': mpl.ratio});\n",
" }\n",
" fig.send_message(\"refresh\", {});\n",
" }\n",
"\n",
" this.imageObj.onload = function() {\n",
" if (fig.image_mode == 'full') {\n",
" // Full images could contain transparency (where diff images\n",
" // almost always do), so we need to clear the canvas so that\n",
" // there is no ghosting.\n",
" fig.context.clearRect(0, 0, fig.canvas.width, fig.canvas.height);\n",
" }\n",
" fig.context.drawImage(fig.imageObj, 0, 0);\n",
" };\n",
"\n",
" this.imageObj.onunload = function() {\n",
" fig.ws.close();\n",
" }\n",
"\n",
" this.ws.onmessage = this._make_on_message_function(this);\n",
"\n",
" this.ondownload = ondownload;\n",
"}\n",
"\n",
"mpl.figure.prototype._init_header = function() {\n",
" var titlebar = $(\n",
" '');\n",
" var titletext = $(\n",
" '');\n",
" titlebar.append(titletext)\n",
" this.root.append(titlebar);\n",
" this.header = titletext[0];\n",
"}\n",
"\n",
"\n",
"\n",
"mpl.figure.prototype._canvas_extra_style = function(canvas_div) {\n",
"\n",
"}\n",
"\n",
"\n",
"mpl.figure.prototype._root_extra_style = function(canvas_div) {\n",
"\n",
"}\n",
"\n",
"mpl.figure.prototype._init_canvas = function() {\n",
" var fig = this;\n",
"\n",
" var canvas_div = $('');\n",
"\n",
" canvas_div.attr('style', 'position: relative; clear: both; outline: 0');\n",
"\n",
" function canvas_keyboard_event(event) {\n",
" return fig.key_event(event, event['data']);\n",
" }\n",
"\n",
" canvas_div.keydown('key_press', canvas_keyboard_event);\n",
" canvas_div.keyup('key_release', canvas_keyboard_event);\n",
" this.canvas_div = canvas_div\n",
" this._canvas_extra_style(canvas_div)\n",
" this.root.append(canvas_div);\n",
"\n",
" var canvas = $('');\n",
" canvas.addClass('mpl-canvas');\n",
" canvas.attr('style', \"left: 0; top: 0; z-index: 0; outline: 0\")\n",
"\n",
" this.canvas = canvas[0];\n",
" this.context = canvas[0].getContext(\"2d\");\n",
"\n",
" var backingStore = this.context.backingStorePixelRatio ||\n",
"\tthis.context.webkitBackingStorePixelRatio ||\n",
"\tthis.context.mozBackingStorePixelRatio ||\n",
"\tthis.context.msBackingStorePixelRatio ||\n",
"\tthis.context.oBackingStorePixelRatio ||\n",
"\tthis.context.backingStorePixelRatio || 1;\n",
"\n",
" mpl.ratio = (window.devicePixelRatio || 1) / backingStore;\n",
"\n",
" var rubberband = $('');\n",
" rubberband.attr('style', \"position: absolute; left: 0; top: 0; z-index: 1;\")\n",
"\n",
" var pass_mouse_events = true;\n",
"\n",
" canvas_div.resizable({\n",
" start: function(event, ui) {\n",
" pass_mouse_events = false;\n",
" },\n",
" resize: function(event, ui) {\n",
" fig.request_resize(ui.size.width, ui.size.height);\n",
" },\n",
" stop: function(event, ui) {\n",
" pass_mouse_events = true;\n",
" fig.request_resize(ui.size.width, ui.size.height);\n",
" },\n",
" });\n",
"\n",
" function mouse_event_fn(event) {\n",
" if (pass_mouse_events)\n",
" return fig.mouse_event(event, event['data']);\n",
" }\n",
"\n",
" rubberband.mousedown('button_press', mouse_event_fn);\n",
" rubberband.mouseup('button_release', mouse_event_fn);\n",
" // Throttle sequential mouse events to 1 every 20ms.\n",
" rubberband.mousemove('motion_notify', mouse_event_fn);\n",
"\n",
" rubberband.mouseenter('figure_enter', mouse_event_fn);\n",
" rubberband.mouseleave('figure_leave', mouse_event_fn);\n",
"\n",
" canvas_div.on(\"wheel\", function (event) {\n",
" event = event.originalEvent;\n",
" event['data'] = 'scroll'\n",
" if (event.deltaY < 0) {\n",
" event.step = 1;\n",
" } else {\n",
" event.step = -1;\n",
" }\n",
" mouse_event_fn(event);\n",
" });\n",
"\n",
" canvas_div.append(canvas);\n",
" canvas_div.append(rubberband);\n",
"\n",
" this.rubberband = rubberband;\n",
" this.rubberband_canvas = rubberband[0];\n",
" this.rubberband_context = rubberband[0].getContext(\"2d\");\n",
" this.rubberband_context.strokeStyle = \"#000000\";\n",
"\n",
" this._resize_canvas = function(width, height) {\n",
" // Keep the size of the canvas, canvas container, and rubber band\n",
" // canvas in synch.\n",
" canvas_div.css('width', width)\n",
" canvas_div.css('height', height)\n",
"\n",
" canvas.attr('width', width * mpl.ratio);\n",
" canvas.attr('height', height * mpl.ratio);\n",
" canvas.attr('style', 'width: ' + width + 'px; height: ' + height + 'px;');\n",
"\n",
" rubberband.attr('width', width);\n",
" rubberband.attr('height', height);\n",
" }\n",
"\n",
" // Set the figure to an initial 600x600px, this will subsequently be updated\n",
" // upon first draw.\n",
" this._resize_canvas(600, 600);\n",
"\n",
" // Disable right mouse context menu.\n",
" $(this.rubberband_canvas).bind(\"contextmenu\",function(e){\n",
" return false;\n",
" });\n",
"\n",
" function set_focus () {\n",
" canvas.focus();\n",
" canvas_div.focus();\n",
" }\n",
"\n",
" window.setTimeout(set_focus, 100);\n",
"}\n",
"\n",
"mpl.figure.prototype._init_toolbar = function() {\n",
" var fig = this;\n",
"\n",
" var nav_element = $('')\n",
" nav_element.attr('style', 'width: 100%');\n",
" this.root.append(nav_element);\n",
"\n",
" // Define a callback function for later on.\n",
" function toolbar_event(event) {\n",
" return fig.toolbar_button_onclick(event['data']);\n",
" }\n",
" function toolbar_mouse_event(event) {\n",
" return fig.toolbar_button_onmouseover(event['data']);\n",
" }\n",
"\n",
" for(var toolbar_ind in mpl.toolbar_items) {\n",
" var name = mpl.toolbar_items[toolbar_ind][0];\n",
" var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
" var image = mpl.toolbar_items[toolbar_ind][2];\n",
" var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
"\n",
" if (!name) {\n",
" // put a spacer in here.\n",
" continue;\n",
" }\n",
" var button = $('');\n",
" button.addClass('ui-button ui-widget ui-state-default ui-corner-all ' +\n",
" 'ui-button-icon-only');\n",
" button.attr('role', 'button');\n",
" button.attr('aria-disabled', 'false');\n",
" button.click(method_name, toolbar_event);\n",
" button.mouseover(tooltip, toolbar_mouse_event);\n",
"\n",
" var icon_img = $('');\n",
" icon_img.addClass('ui-button-icon-primary ui-icon');\n",
" icon_img.addClass(image);\n",
" icon_img.addClass('ui-corner-all');\n",
"\n",
" var tooltip_span = $('');\n",
" tooltip_span.addClass('ui-button-text');\n",
" tooltip_span.html(tooltip);\n",
"\n",
" button.append(icon_img);\n",
" button.append(tooltip_span);\n",
"\n",
" nav_element.append(button);\n",
" }\n",
"\n",
" var fmt_picker_span = $('');\n",
"\n",
" var fmt_picker = $('');\n",
" fmt_picker.addClass('mpl-toolbar-option ui-widget ui-widget-content');\n",
" fmt_picker_span.append(fmt_picker);\n",
" nav_element.append(fmt_picker_span);\n",
" this.format_dropdown = fmt_picker[0];\n",
"\n",
" for (var ind in mpl.extensions) {\n",
" var fmt = mpl.extensions[ind];\n",
" var option = $(\n",
" '', {selected: fmt === mpl.default_extension}).html(fmt);\n",
" fmt_picker.append(option)\n",
" }\n",
"\n",
" // Add hover states to the ui-buttons\n",
" $( \".ui-button\" ).hover(\n",
" function() { $(this).addClass(\"ui-state-hover\");},\n",
" function() { $(this).removeClass(\"ui-state-hover\");}\n",
" );\n",
"\n",
" var status_bar = $('');\n",
" nav_element.append(status_bar);\n",
" this.message = status_bar[0];\n",
"}\n",
"\n",
"mpl.figure.prototype.request_resize = function(x_pixels, y_pixels) {\n",
" // Request matplotlib to resize the figure. Matplotlib will then trigger a resize in the client,\n",
" // which will in turn request a refresh of the image.\n",
" this.send_message('resize', {'width': x_pixels, 'height': y_pixels});\n",
"}\n",
"\n",
"mpl.figure.prototype.send_message = function(type, properties) {\n",
" properties['type'] = type;\n",
" properties['figure_id'] = this.id;\n",
" this.ws.send(JSON.stringify(properties));\n",
"}\n",
"\n",
"mpl.figure.prototype.send_draw_message = function() {\n",
" if (!this.waiting) {\n",
" this.waiting = true;\n",
" this.ws.send(JSON.stringify({type: \"draw\", figure_id: this.id}));\n",
" }\n",
"}\n",
"\n",
"\n",
"mpl.figure.prototype.handle_save = function(fig, msg) {\n",
" var format_dropdown = fig.format_dropdown;\n",
" var format = format_dropdown.options[format_dropdown.selectedIndex].value;\n",
" fig.ondownload(fig, format);\n",
"}\n",
"\n",
"\n",
"mpl.figure.prototype.handle_resize = function(fig, msg) {\n",
" var size = msg['size'];\n",
" if (size[0] != fig.canvas.width || size[1] != fig.canvas.height) {\n",
" fig._resize_canvas(size[0], size[1]);\n",
" fig.send_message(\"refresh\", {});\n",
" };\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_rubberband = function(fig, msg) {\n",
" var x0 = msg['x0'] / mpl.ratio;\n",
" var y0 = (fig.canvas.height - msg['y0']) / mpl.ratio;\n",
" var x1 = msg['x1'] / mpl.ratio;\n",
" var y1 = (fig.canvas.height - msg['y1']) / mpl.ratio;\n",
" x0 = Math.floor(x0) + 0.5;\n",
" y0 = Math.floor(y0) + 0.5;\n",
" x1 = Math.floor(x1) + 0.5;\n",
" y1 = Math.floor(y1) + 0.5;\n",
" var min_x = Math.min(x0, x1);\n",
" var min_y = Math.min(y0, y1);\n",
" var width = Math.abs(x1 - x0);\n",
" var height = Math.abs(y1 - y0);\n",
"\n",
" fig.rubberband_context.clearRect(\n",
" 0, 0, fig.canvas.width, fig.canvas.height);\n",
"\n",
" fig.rubberband_context.strokeRect(min_x, min_y, width, height);\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_figure_label = function(fig, msg) {\n",
" // Updates the figure title.\n",
" fig.header.textContent = msg['label'];\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_cursor = function(fig, msg) {\n",
" var cursor = msg['cursor'];\n",
" switch(cursor)\n",
" {\n",
" case 0:\n",
" cursor = 'pointer';\n",
" break;\n",
" case 1:\n",
" cursor = 'default';\n",
" break;\n",
" case 2:\n",
" cursor = 'crosshair';\n",
" break;\n",
" case 3:\n",
" cursor = 'move';\n",
" break;\n",
" }\n",
" fig.rubberband_canvas.style.cursor = cursor;\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_message = function(fig, msg) {\n",
" fig.message.textContent = msg['message'];\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_draw = function(fig, msg) {\n",
" // Request the server to send over a new figure.\n",
" fig.send_draw_message();\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_image_mode = function(fig, msg) {\n",
" fig.image_mode = msg['mode'];\n",
"}\n",
"\n",
"mpl.figure.prototype.updated_canvas_event = function() {\n",
" // Called whenever the canvas gets updated.\n",
" this.send_message(\"ack\", {});\n",
"}\n",
"\n",
"// A function to construct a web socket function for onmessage handling.\n",
"// Called in the figure constructor.\n",
"mpl.figure.prototype._make_on_message_function = function(fig) {\n",
" return function socket_on_message(evt) {\n",
" if (evt.data instanceof Blob) {\n",
" /* FIXME: We get \"Resource interpreted as Image but\n",
" * transferred with MIME type text/plain:\" errors on\n",
" * Chrome. But how to set the MIME type? It doesn't seem\n",
" * to be part of the websocket stream */\n",
" evt.data.type = \"image/png\";\n",
"\n",
" /* Free the memory for the previous frames */\n",
" if (fig.imageObj.src) {\n",
" (window.URL || window.webkitURL).revokeObjectURL(\n",
" fig.imageObj.src);\n",
" }\n",
"\n",
" fig.imageObj.src = (window.URL || window.webkitURL).createObjectURL(\n",
" evt.data);\n",
" fig.updated_canvas_event();\n",
" fig.waiting = false;\n",
" return;\n",
" }\n",
" else if (typeof evt.data === 'string' && evt.data.slice(0, 21) == \"data:image/png;base64\") {\n",
" fig.imageObj.src = evt.data;\n",
" fig.updated_canvas_event();\n",
" fig.waiting = false;\n",
" return;\n",
" }\n",
"\n",
" var msg = JSON.parse(evt.data);\n",
" var msg_type = msg['type'];\n",
"\n",
" // Call the \"handle_{type}\" callback, which takes\n",
" // the figure and JSON message as its only arguments.\n",
" try {\n",
" var callback = fig[\"handle_\" + msg_type];\n",
" } catch (e) {\n",
" console.log(\"No handler for the '\" + msg_type + \"' message type: \", msg);\n",
" return;\n",
" }\n",
"\n",
" if (callback) {\n",
" try {\n",
" // console.log(\"Handling '\" + msg_type + \"' message: \", msg);\n",
" callback(fig, msg);\n",
" } catch (e) {\n",
" console.log(\"Exception inside the 'handler_\" + msg_type + \"' callback:\", e, e.stack, msg);\n",
" }\n",
" }\n",
" };\n",
"}\n",
"\n",
"// from http://stackoverflow.com/questions/1114465/getting-mouse-location-in-canvas\n",
"mpl.findpos = function(e) {\n",
" //this section is from http://www.quirksmode.org/js/events_properties.html\n",
" var targ;\n",
" if (!e)\n",
" e = window.event;\n",
" if (e.target)\n",
" targ = e.target;\n",
" else if (e.srcElement)\n",
" targ = e.srcElement;\n",
" if (targ.nodeType == 3) // defeat Safari bug\n",
" targ = targ.parentNode;\n",
"\n",
" // jQuery normalizes the pageX and pageY\n",
" // pageX,Y are the mouse positions relative to the document\n",
" // offset() returns the position of the element relative to the document\n",
" var x = e.pageX - $(targ).offset().left;\n",
" var y = e.pageY - $(targ).offset().top;\n",
"\n",
" return {\"x\": x, \"y\": y};\n",
"};\n",
"\n",
"/*\n",
" * return a copy of an object with only non-object keys\n",
" * we need this to avoid circular references\n",
" * http://stackoverflow.com/a/24161582/3208463\n",
" */\n",
"function simpleKeys (original) {\n",
" return Object.keys(original).reduce(function (obj, key) {\n",
" if (typeof original[key] !== 'object')\n",
" obj[key] = original[key]\n",
" return obj;\n",
" }, {});\n",
"}\n",
"\n",
"mpl.figure.prototype.mouse_event = function(event, name) {\n",
" var canvas_pos = mpl.findpos(event)\n",
"\n",
" if (name === 'button_press')\n",
" {\n",
" this.canvas.focus();\n",
" this.canvas_div.focus();\n",
" }\n",
"\n",
" var x = canvas_pos.x * mpl.ratio;\n",
" var y = canvas_pos.y * mpl.ratio;\n",
"\n",
" this.send_message(name, {x: x, y: y, button: event.button,\n",
" step: event.step,\n",
" guiEvent: simpleKeys(event)});\n",
"\n",
" /* This prevents the web browser from automatically changing to\n",
" * the text insertion cursor when the button is pressed. We want\n",
" * to control all of the cursor setting manually through the\n",
" * 'cursor' event from matplotlib */\n",
" event.preventDefault();\n",
" return false;\n",
"}\n",
"\n",
"mpl.figure.prototype._key_event_extra = function(event, name) {\n",
" // Handle any extra behaviour associated with a key event\n",
"}\n",
"\n",
"mpl.figure.prototype.key_event = function(event, name) {\n",
"\n",
" // Prevent repeat events\n",
" if (name == 'key_press')\n",
" {\n",
" if (event.which === this._key)\n",
" return;\n",
" else\n",
" this._key = event.which;\n",
" }\n",
" if (name == 'key_release')\n",
" this._key = null;\n",
"\n",
" var value = '';\n",
" if (event.ctrlKey && event.which != 17)\n",
" value += \"ctrl+\";\n",
" if (event.altKey && event.which != 18)\n",
" value += \"alt+\";\n",
" if (event.shiftKey && event.which != 16)\n",
" value += \"shift+\";\n",
"\n",
" value += 'k';\n",
" value += event.which.toString();\n",
"\n",
" this._key_event_extra(event, name);\n",
"\n",
" this.send_message(name, {key: value,\n",
" guiEvent: simpleKeys(event)});\n",
" return false;\n",
"}\n",
"\n",
"mpl.figure.prototype.toolbar_button_onclick = function(name) {\n",
" if (name == 'download') {\n",
" this.handle_save(this, null);\n",
" } else {\n",
" this.send_message(\"toolbar_button\", {name: name});\n",
" }\n",
"};\n",
"\n",
"mpl.figure.prototype.toolbar_button_onmouseover = function(tooltip) {\n",
" this.message.textContent = tooltip;\n",
"};\n",
"mpl.toolbar_items = [[\"Home\", \"Reset original view\", \"fa fa-home icon-home\", \"home\"], [\"Back\", \"Back to previous view\", \"fa fa-arrow-left icon-arrow-left\", \"back\"], [\"Forward\", \"Forward to next view\", \"fa fa-arrow-right icon-arrow-right\", \"forward\"], [\"\", \"\", \"\", \"\"], [\"Pan\", \"Pan axes with left mouse, zoom with right\", \"fa fa-arrows icon-move\", \"pan\"], [\"Zoom\", \"Zoom to rectangle\", \"fa fa-square-o icon-check-empty\", \"zoom\"], [\"\", \"\", \"\", \"\"], [\"Download\", \"Download plot\", \"fa fa-floppy-o icon-save\", \"download\"]];\n",
"\n",
"mpl.extensions = [\"eps\", \"jpeg\", \"pdf\", \"png\", \"ps\", \"raw\", \"svg\", \"tif\"];\n",
"\n",
"mpl.default_extension = \"png\";var comm_websocket_adapter = function(comm) {\n",
" // Create a \"websocket\"-like object which calls the given IPython comm\n",
" // object with the appropriate methods. Currently this is a non binary\n",
" // socket, so there is still some room for performance tuning.\n",
" var ws = {};\n",
"\n",
" ws.close = function() {\n",
" comm.close()\n",
" };\n",
" ws.send = function(m) {\n",
" //console.log('sending', m);\n",
" comm.send(m);\n",
" };\n",
" // Register the callback with on_msg.\n",
" comm.on_msg(function(msg) {\n",
" //console.log('receiving', msg['content']['data'], msg);\n",
" // Pass the mpl event to the overridden (by mpl) onmessage function.\n",
" ws.onmessage(msg['content']['data'])\n",
" });\n",
" return ws;\n",
"}\n",
"\n",
"mpl.mpl_figure_comm = function(comm, msg) {\n",
" // This is the function which gets called when the mpl process\n",
" // starts-up an IPython Comm through the \"matplotlib\" channel.\n",
"\n",
" var id = msg.content.data.id;\n",
" // Get hold of the div created by the display call when the Comm\n",
" // socket was opened in Python.\n",
" var element = $(\"#\" + id);\n",
" var ws_proxy = comm_websocket_adapter(comm)\n",
"\n",
" function ondownload(figure, format) {\n",
" window.open(figure.imageObj.src);\n",
" }\n",
"\n",
" var fig = new mpl.figure(id, ws_proxy,\n",
" ondownload,\n",
" element.get(0));\n",
"\n",
" // Call onopen now - mpl needs it, as it is assuming we've passed it a real\n",
" // web socket which is closed, not our websocket->open comm proxy.\n",
" ws_proxy.onopen();\n",
"\n",
" fig.parent_element = element.get(0);\n",
" fig.cell_info = mpl.find_output_cell(\"\");\n",
" if (!fig.cell_info) {\n",
" console.error(\"Failed to find cell for figure\", id, fig);\n",
" return;\n",
" }\n",
"\n",
" var output_index = fig.cell_info[2]\n",
" var cell = fig.cell_info[0];\n",
"\n",
"};\n",
"\n",
"mpl.figure.prototype.handle_close = function(fig, msg) {\n",
" var width = fig.canvas.width/mpl.ratio\n",
" fig.root.unbind('remove')\n",
"\n",
" // Update the output cell to use the data from the current canvas.\n",
" fig.push_to_output();\n",
" var dataURL = fig.canvas.toDataURL();\n",
" // Re-enable the keyboard manager in IPython - without this line, in FF,\n",
" // the notebook keyboard shortcuts fail.\n",
" IPython.keyboard_manager.enable()\n",
" $(fig.parent_element).html('');\n",
" fig.close_ws(fig, msg);\n",
"}\n",
"\n",
"mpl.figure.prototype.close_ws = function(fig, msg){\n",
" fig.send_message('closing', msg);\n",
" // fig.ws.close()\n",
"}\n",
"\n",
"mpl.figure.prototype.push_to_output = function(remove_interactive) {\n",
" // Turn the data on the canvas into data in the output cell.\n",
" var width = this.canvas.width/mpl.ratio\n",
" var dataURL = this.canvas.toDataURL();\n",
" this.cell_info[1]['text/html'] = '';\n",
"}\n",
"\n",
"mpl.figure.prototype.updated_canvas_event = function() {\n",
" // Tell IPython that the notebook contents must change.\n",
" IPython.notebook.set_dirty(true);\n",
" this.send_message(\"ack\", {});\n",
" var fig = this;\n",
" // Wait a second, then push the new image to the DOM so\n",
" // that it is saved nicely (might be nice to debounce this).\n",
" setTimeout(function () { fig.push_to_output() }, 1000);\n",
"}\n",
"\n",
"mpl.figure.prototype._init_toolbar = function() {\n",
" var fig = this;\n",
"\n",
" var nav_element = $('')\n",
" nav_element.attr('style', 'width: 100%');\n",
" this.root.append(nav_element);\n",
"\n",
" // Define a callback function for later on.\n",
" function toolbar_event(event) {\n",
" return fig.toolbar_button_onclick(event['data']);\n",
" }\n",
" function toolbar_mouse_event(event) {\n",
" return fig.toolbar_button_onmouseover(event['data']);\n",
" }\n",
"\n",
" for(var toolbar_ind in mpl.toolbar_items){\n",
" var name = mpl.toolbar_items[toolbar_ind][0];\n",
" var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
" var image = mpl.toolbar_items[toolbar_ind][2];\n",
" var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
"\n",
" if (!name) { continue; };\n",
"\n",
" var button = $('');\n",
" button.click(method_name, toolbar_event);\n",
" button.mouseover(tooltip, toolbar_mouse_event);\n",
" nav_element.append(button);\n",
" }\n",
"\n",
" // Add the status bar.\n",
" var status_bar = $('');\n",
" nav_element.append(status_bar);\n",
" this.message = status_bar[0];\n",
"\n",
" // Add the close button to the window.\n",
" var buttongrp = $('');\n",
" var button = $('');\n",
" button.click(function (evt) { fig.handle_close(fig, {}); } );\n",
" button.mouseover('Stop Interaction', toolbar_mouse_event);\n",
" buttongrp.append(button);\n",
" var titlebar = this.root.find($('.ui-dialog-titlebar'));\n",
" titlebar.prepend(buttongrp);\n",
"}\n",
"\n",
"mpl.figure.prototype._root_extra_style = function(el){\n",
" var fig = this\n",
" el.on(\"remove\", function(){\n",
"\tfig.close_ws(fig, {});\n",
" });\n",
"}\n",
"\n",
"mpl.figure.prototype._canvas_extra_style = function(el){\n",
" // this is important to make the div 'focusable\n",
" el.attr('tabindex', 0)\n",
" // reach out to IPython and tell the keyboard manager to turn it's self\n",
" // off when our div gets focus\n",
"\n",
" // location in version 3\n",
" if (IPython.notebook.keyboard_manager) {\n",
" IPython.notebook.keyboard_manager.register_events(el);\n",
" }\n",
" else {\n",
" // location in version 2\n",
" IPython.keyboard_manager.register_events(el);\n",
" }\n",
"\n",
"}\n",
"\n",
"mpl.figure.prototype._key_event_extra = function(event, name) {\n",
" var manager = IPython.notebook.keyboard_manager;\n",
" if (!manager)\n",
" manager = IPython.keyboard_manager;\n",
"\n",
" // Check for shift+enter\n",
" if (event.shiftKey && event.which == 13) {\n",
" this.canvas_div.blur();\n",
" event.shiftKey = false;\n",
" // Send a \"J\" for go to next cell\n",
" event.which = 74;\n",
" event.keyCode = 74;\n",
" manager.command_mode();\n",
" manager.handle_keydown(event);\n",
" }\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_save = function(fig, msg) {\n",
" fig.ondownload(fig, null);\n",
"}\n",
"\n",
"\n",
"mpl.find_output_cell = function(html_output) {\n",
" // Return the cell and output element which can be found *uniquely* in the notebook.\n",
" // Note - this is a bit hacky, but it is done because the \"notebook_saving.Notebook\"\n",
" // IPython event is triggered only after the cells have been serialised, which for\n",
" // our purposes (turning an active figure into a static one), is too late.\n",
" var cells = IPython.notebook.get_cells();\n",
" var ncells = cells.length;\n",
" for (var i=0; i= 3 moved mimebundle to data attribute of output\n",
" data = data.data;\n",
" }\n",
" if (data['text/html'] == html_output) {\n",
" return [cell, data, j];\n",
" }\n",
" }\n",
" }\n",
" }\n",
"}\n",
"\n",
"// Register the function which deals with the matplotlib target/channel.\n",
"// The kernel may be null if the page has been refreshed.\n",
"if (IPython.notebook.kernel != null) {\n",
" IPython.notebook.kernel.comm_manager.register_target('matplotlib', mpl.mpl_figure_comm);\n",
"}\n"
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/html": [
""
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/javascript": [
"/* Put everything inside the global mpl namespace */\n",
"window.mpl = {};\n",
"\n",
"\n",
"mpl.get_websocket_type = function() {\n",
" if (typeof(WebSocket) !== 'undefined') {\n",
" return WebSocket;\n",
" } else if (typeof(MozWebSocket) !== 'undefined') {\n",
" return MozWebSocket;\n",
" } else {\n",
" alert('Your browser does not have WebSocket support.' +\n",
" 'Please try Chrome, Safari or Firefox ≥ 6. ' +\n",
" 'Firefox 4 and 5 are also supported but you ' +\n",
" 'have to enable WebSockets in about:config.');\n",
" };\n",
"}\n",
"\n",
"mpl.figure = function(figure_id, websocket, ondownload, parent_element) {\n",
" this.id = figure_id;\n",
"\n",
" this.ws = websocket;\n",
"\n",
" this.supports_binary = (this.ws.binaryType != undefined);\n",
"\n",
" if (!this.supports_binary) {\n",
" var warnings = document.getElementById(\"mpl-warnings\");\n",
" if (warnings) {\n",
" warnings.style.display = 'block';\n",
" warnings.textContent = (\n",
" \"This browser does not support binary websocket messages. \" +\n",
" \"Performance may be slow.\");\n",
" }\n",
" }\n",
"\n",
" this.imageObj = new Image();\n",
"\n",
" this.context = undefined;\n",
" this.message = undefined;\n",
" this.canvas = undefined;\n",
" this.rubberband_canvas = undefined;\n",
" this.rubberband_context = undefined;\n",
" this.format_dropdown = undefined;\n",
"\n",
" this.image_mode = 'full';\n",
"\n",
" this.root = $('');\n",
" this._root_extra_style(this.root)\n",
" this.root.attr('style', 'display: inline-block');\n",
"\n",
" $(parent_element).append(this.root);\n",
"\n",
" this._init_header(this);\n",
" this._init_canvas(this);\n",
" this._init_toolbar(this);\n",
"\n",
" var fig = this;\n",
"\n",
" this.waiting = false;\n",
"\n",
" this.ws.onopen = function () {\n",
" fig.send_message(\"supports_binary\", {value: fig.supports_binary});\n",
" fig.send_message(\"send_image_mode\", {});\n",
" if (mpl.ratio != 1) {\n",
" fig.send_message(\"set_dpi_ratio\", {'dpi_ratio': mpl.ratio});\n",
" }\n",
" fig.send_message(\"refresh\", {});\n",
" }\n",
"\n",
" this.imageObj.onload = function() {\n",
" if (fig.image_mode == 'full') {\n",
" // Full images could contain transparency (where diff images\n",
" // almost always do), so we need to clear the canvas so that\n",
" // there is no ghosting.\n",
" fig.context.clearRect(0, 0, fig.canvas.width, fig.canvas.height);\n",
" }\n",
" fig.context.drawImage(fig.imageObj, 0, 0);\n",
" };\n",
"\n",
" this.imageObj.onunload = function() {\n",
" fig.ws.close();\n",
" }\n",
"\n",
" this.ws.onmessage = this._make_on_message_function(this);\n",
"\n",
" this.ondownload = ondownload;\n",
"}\n",
"\n",
"mpl.figure.prototype._init_header = function() {\n",
" var titlebar = $(\n",
" '');\n",
" var titletext = $(\n",
" '');\n",
" titlebar.append(titletext)\n",
" this.root.append(titlebar);\n",
" this.header = titletext[0];\n",
"}\n",
"\n",
"\n",
"\n",
"mpl.figure.prototype._canvas_extra_style = function(canvas_div) {\n",
"\n",
"}\n",
"\n",
"\n",
"mpl.figure.prototype._root_extra_style = function(canvas_div) {\n",
"\n",
"}\n",
"\n",
"mpl.figure.prototype._init_canvas = function() {\n",
" var fig = this;\n",
"\n",
" var canvas_div = $('');\n",
"\n",
" canvas_div.attr('style', 'position: relative; clear: both; outline: 0');\n",
"\n",
" function canvas_keyboard_event(event) {\n",
" return fig.key_event(event, event['data']);\n",
" }\n",
"\n",
" canvas_div.keydown('key_press', canvas_keyboard_event);\n",
" canvas_div.keyup('key_release', canvas_keyboard_event);\n",
" this.canvas_div = canvas_div\n",
" this._canvas_extra_style(canvas_div)\n",
" this.root.append(canvas_div);\n",
"\n",
" var canvas = $('');\n",
" canvas.addClass('mpl-canvas');\n",
" canvas.attr('style', \"left: 0; top: 0; z-index: 0; outline: 0\")\n",
"\n",
" this.canvas = canvas[0];\n",
" this.context = canvas[0].getContext(\"2d\");\n",
"\n",
" var backingStore = this.context.backingStorePixelRatio ||\n",
"\tthis.context.webkitBackingStorePixelRatio ||\n",
"\tthis.context.mozBackingStorePixelRatio ||\n",
"\tthis.context.msBackingStorePixelRatio ||\n",
"\tthis.context.oBackingStorePixelRatio ||\n",
"\tthis.context.backingStorePixelRatio || 1;\n",
"\n",
" mpl.ratio = (window.devicePixelRatio || 1) / backingStore;\n",
"\n",
" var rubberband = $('');\n",
" rubberband.attr('style', \"position: absolute; left: 0; top: 0; z-index: 1;\")\n",
"\n",
" var pass_mouse_events = true;\n",
"\n",
" canvas_div.resizable({\n",
" start: function(event, ui) {\n",
" pass_mouse_events = false;\n",
" },\n",
" resize: function(event, ui) {\n",
" fig.request_resize(ui.size.width, ui.size.height);\n",
" },\n",
" stop: function(event, ui) {\n",
" pass_mouse_events = true;\n",
" fig.request_resize(ui.size.width, ui.size.height);\n",
" },\n",
" });\n",
"\n",
" function mouse_event_fn(event) {\n",
" if (pass_mouse_events)\n",
" return fig.mouse_event(event, event['data']);\n",
" }\n",
"\n",
" rubberband.mousedown('button_press', mouse_event_fn);\n",
" rubberband.mouseup('button_release', mouse_event_fn);\n",
" // Throttle sequential mouse events to 1 every 20ms.\n",
" rubberband.mousemove('motion_notify', mouse_event_fn);\n",
"\n",
" rubberband.mouseenter('figure_enter', mouse_event_fn);\n",
" rubberband.mouseleave('figure_leave', mouse_event_fn);\n",
"\n",
" canvas_div.on(\"wheel\", function (event) {\n",
" event = event.originalEvent;\n",
" event['data'] = 'scroll'\n",
" if (event.deltaY < 0) {\n",
" event.step = 1;\n",
" } else {\n",
" event.step = -1;\n",
" }\n",
" mouse_event_fn(event);\n",
" });\n",
"\n",
" canvas_div.append(canvas);\n",
" canvas_div.append(rubberband);\n",
"\n",
" this.rubberband = rubberband;\n",
" this.rubberband_canvas = rubberband[0];\n",
" this.rubberband_context = rubberband[0].getContext(\"2d\");\n",
" this.rubberband_context.strokeStyle = \"#000000\";\n",
"\n",
" this._resize_canvas = function(width, height) {\n",
" // Keep the size of the canvas, canvas container, and rubber band\n",
" // canvas in synch.\n",
" canvas_div.css('width', width)\n",
" canvas_div.css('height', height)\n",
"\n",
" canvas.attr('width', width * mpl.ratio);\n",
" canvas.attr('height', height * mpl.ratio);\n",
" canvas.attr('style', 'width: ' + width + 'px; height: ' + height + 'px;');\n",
"\n",
" rubberband.attr('width', width);\n",
" rubberband.attr('height', height);\n",
" }\n",
"\n",
" // Set the figure to an initial 600x600px, this will subsequently be updated\n",
" // upon first draw.\n",
" this._resize_canvas(600, 600);\n",
"\n",
" // Disable right mouse context menu.\n",
" $(this.rubberband_canvas).bind(\"contextmenu\",function(e){\n",
" return false;\n",
" });\n",
"\n",
" function set_focus () {\n",
" canvas.focus();\n",
" canvas_div.focus();\n",
" }\n",
"\n",
" window.setTimeout(set_focus, 100);\n",
"}\n",
"\n",
"mpl.figure.prototype._init_toolbar = function() {\n",
" var fig = this;\n",
"\n",
" var nav_element = $('')\n",
" nav_element.attr('style', 'width: 100%');\n",
" this.root.append(nav_element);\n",
"\n",
" // Define a callback function for later on.\n",
" function toolbar_event(event) {\n",
" return fig.toolbar_button_onclick(event['data']);\n",
" }\n",
" function toolbar_mouse_event(event) {\n",
" return fig.toolbar_button_onmouseover(event['data']);\n",
" }\n",
"\n",
" for(var toolbar_ind in mpl.toolbar_items) {\n",
" var name = mpl.toolbar_items[toolbar_ind][0];\n",
" var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
" var image = mpl.toolbar_items[toolbar_ind][2];\n",
" var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
"\n",
" if (!name) {\n",
" // put a spacer in here.\n",
" continue;\n",
" }\n",
" var button = $('');\n",
" button.addClass('ui-button ui-widget ui-state-default ui-corner-all ' +\n",
" 'ui-button-icon-only');\n",
" button.attr('role', 'button');\n",
" button.attr('aria-disabled', 'false');\n",
" button.click(method_name, toolbar_event);\n",
" button.mouseover(tooltip, toolbar_mouse_event);\n",
"\n",
" var icon_img = $('');\n",
" icon_img.addClass('ui-button-icon-primary ui-icon');\n",
" icon_img.addClass(image);\n",
" icon_img.addClass('ui-corner-all');\n",
"\n",
" var tooltip_span = $('');\n",
" tooltip_span.addClass('ui-button-text');\n",
" tooltip_span.html(tooltip);\n",
"\n",
" button.append(icon_img);\n",
" button.append(tooltip_span);\n",
"\n",
" nav_element.append(button);\n",
" }\n",
"\n",
" var fmt_picker_span = $('');\n",
"\n",
" var fmt_picker = $('');\n",
" fmt_picker.addClass('mpl-toolbar-option ui-widget ui-widget-content');\n",
" fmt_picker_span.append(fmt_picker);\n",
" nav_element.append(fmt_picker_span);\n",
" this.format_dropdown = fmt_picker[0];\n",
"\n",
" for (var ind in mpl.extensions) {\n",
" var fmt = mpl.extensions[ind];\n",
" var option = $(\n",
" '', {selected: fmt === mpl.default_extension}).html(fmt);\n",
" fmt_picker.append(option)\n",
" }\n",
"\n",
" // Add hover states to the ui-buttons\n",
" $( \".ui-button\" ).hover(\n",
" function() { $(this).addClass(\"ui-state-hover\");},\n",
" function() { $(this).removeClass(\"ui-state-hover\");}\n",
" );\n",
"\n",
" var status_bar = $('');\n",
" nav_element.append(status_bar);\n",
" this.message = status_bar[0];\n",
"}\n",
"\n",
"mpl.figure.prototype.request_resize = function(x_pixels, y_pixels) {\n",
" // Request matplotlib to resize the figure. Matplotlib will then trigger a resize in the client,\n",
" // which will in turn request a refresh of the image.\n",
" this.send_message('resize', {'width': x_pixels, 'height': y_pixels});\n",
"}\n",
"\n",
"mpl.figure.prototype.send_message = function(type, properties) {\n",
" properties['type'] = type;\n",
" properties['figure_id'] = this.id;\n",
" this.ws.send(JSON.stringify(properties));\n",
"}\n",
"\n",
"mpl.figure.prototype.send_draw_message = function() {\n",
" if (!this.waiting) {\n",
" this.waiting = true;\n",
" this.ws.send(JSON.stringify({type: \"draw\", figure_id: this.id}));\n",
" }\n",
"}\n",
"\n",
"\n",
"mpl.figure.prototype.handle_save = function(fig, msg) {\n",
" var format_dropdown = fig.format_dropdown;\n",
" var format = format_dropdown.options[format_dropdown.selectedIndex].value;\n",
" fig.ondownload(fig, format);\n",
"}\n",
"\n",
"\n",
"mpl.figure.prototype.handle_resize = function(fig, msg) {\n",
" var size = msg['size'];\n",
" if (size[0] != fig.canvas.width || size[1] != fig.canvas.height) {\n",
" fig._resize_canvas(size[0], size[1]);\n",
" fig.send_message(\"refresh\", {});\n",
" };\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_rubberband = function(fig, msg) {\n",
" var x0 = msg['x0'] / mpl.ratio;\n",
" var y0 = (fig.canvas.height - msg['y0']) / mpl.ratio;\n",
" var x1 = msg['x1'] / mpl.ratio;\n",
" var y1 = (fig.canvas.height - msg['y1']) / mpl.ratio;\n",
" x0 = Math.floor(x0) + 0.5;\n",
" y0 = Math.floor(y0) + 0.5;\n",
" x1 = Math.floor(x1) + 0.5;\n",
" y1 = Math.floor(y1) + 0.5;\n",
" var min_x = Math.min(x0, x1);\n",
" var min_y = Math.min(y0, y1);\n",
" var width = Math.abs(x1 - x0);\n",
" var height = Math.abs(y1 - y0);\n",
"\n",
" fig.rubberband_context.clearRect(\n",
" 0, 0, fig.canvas.width, fig.canvas.height);\n",
"\n",
" fig.rubberband_context.strokeRect(min_x, min_y, width, height);\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_figure_label = function(fig, msg) {\n",
" // Updates the figure title.\n",
" fig.header.textContent = msg['label'];\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_cursor = function(fig, msg) {\n",
" var cursor = msg['cursor'];\n",
" switch(cursor)\n",
" {\n",
" case 0:\n",
" cursor = 'pointer';\n",
" break;\n",
" case 1:\n",
" cursor = 'default';\n",
" break;\n",
" case 2:\n",
" cursor = 'crosshair';\n",
" break;\n",
" case 3:\n",
" cursor = 'move';\n",
" break;\n",
" }\n",
" fig.rubberband_canvas.style.cursor = cursor;\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_message = function(fig, msg) {\n",
" fig.message.textContent = msg['message'];\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_draw = function(fig, msg) {\n",
" // Request the server to send over a new figure.\n",
" fig.send_draw_message();\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_image_mode = function(fig, msg) {\n",
" fig.image_mode = msg['mode'];\n",
"}\n",
"\n",
"mpl.figure.prototype.updated_canvas_event = function() {\n",
" // Called whenever the canvas gets updated.\n",
" this.send_message(\"ack\", {});\n",
"}\n",
"\n",
"// A function to construct a web socket function for onmessage handling.\n",
"// Called in the figure constructor.\n",
"mpl.figure.prototype._make_on_message_function = function(fig) {\n",
" return function socket_on_message(evt) {\n",
" if (evt.data instanceof Blob) {\n",
" /* FIXME: We get \"Resource interpreted as Image but\n",
" * transferred with MIME type text/plain:\" errors on\n",
" * Chrome. But how to set the MIME type? It doesn't seem\n",
" * to be part of the websocket stream */\n",
" evt.data.type = \"image/png\";\n",
"\n",
" /* Free the memory for the previous frames */\n",
" if (fig.imageObj.src) {\n",
" (window.URL || window.webkitURL).revokeObjectURL(\n",
" fig.imageObj.src);\n",
" }\n",
"\n",
" fig.imageObj.src = (window.URL || window.webkitURL).createObjectURL(\n",
" evt.data);\n",
" fig.updated_canvas_event();\n",
" fig.waiting = false;\n",
" return;\n",
" }\n",
" else if (typeof evt.data === 'string' && evt.data.slice(0, 21) == \"data:image/png;base64\") {\n",
" fig.imageObj.src = evt.data;\n",
" fig.updated_canvas_event();\n",
" fig.waiting = false;\n",
" return;\n",
" }\n",
"\n",
" var msg = JSON.parse(evt.data);\n",
" var msg_type = msg['type'];\n",
"\n",
" // Call the \"handle_{type}\" callback, which takes\n",
" // the figure and JSON message as its only arguments.\n",
" try {\n",
" var callback = fig[\"handle_\" + msg_type];\n",
" } catch (e) {\n",
" console.log(\"No handler for the '\" + msg_type + \"' message type: \", msg);\n",
" return;\n",
" }\n",
"\n",
" if (callback) {\n",
" try {\n",
" // console.log(\"Handling '\" + msg_type + \"' message: \", msg);\n",
" callback(fig, msg);\n",
" } catch (e) {\n",
" console.log(\"Exception inside the 'handler_\" + msg_type + \"' callback:\", e, e.stack, msg);\n",
" }\n",
" }\n",
" };\n",
"}\n",
"\n",
"// from http://stackoverflow.com/questions/1114465/getting-mouse-location-in-canvas\n",
"mpl.findpos = function(e) {\n",
" //this section is from http://www.quirksmode.org/js/events_properties.html\n",
" var targ;\n",
" if (!e)\n",
" e = window.event;\n",
" if (e.target)\n",
" targ = e.target;\n",
" else if (e.srcElement)\n",
" targ = e.srcElement;\n",
" if (targ.nodeType == 3) // defeat Safari bug\n",
" targ = targ.parentNode;\n",
"\n",
" // jQuery normalizes the pageX and pageY\n",
" // pageX,Y are the mouse positions relative to the document\n",
" // offset() returns the position of the element relative to the document\n",
" var x = e.pageX - $(targ).offset().left;\n",
" var y = e.pageY - $(targ).offset().top;\n",
"\n",
" return {\"x\": x, \"y\": y};\n",
"};\n",
"\n",
"/*\n",
" * return a copy of an object with only non-object keys\n",
" * we need this to avoid circular references\n",
" * http://stackoverflow.com/a/24161582/3208463\n",
" */\n",
"function simpleKeys (original) {\n",
" return Object.keys(original).reduce(function (obj, key) {\n",
" if (typeof original[key] !== 'object')\n",
" obj[key] = original[key]\n",
" return obj;\n",
" }, {});\n",
"}\n",
"\n",
"mpl.figure.prototype.mouse_event = function(event, name) {\n",
" var canvas_pos = mpl.findpos(event)\n",
"\n",
" if (name === 'button_press')\n",
" {\n",
" this.canvas.focus();\n",
" this.canvas_div.focus();\n",
" }\n",
"\n",
" var x = canvas_pos.x * mpl.ratio;\n",
" var y = canvas_pos.y * mpl.ratio;\n",
"\n",
" this.send_message(name, {x: x, y: y, button: event.button,\n",
" step: event.step,\n",
" guiEvent: simpleKeys(event)});\n",
"\n",
" /* This prevents the web browser from automatically changing to\n",
" * the text insertion cursor when the button is pressed. We want\n",
" * to control all of the cursor setting manually through the\n",
" * 'cursor' event from matplotlib */\n",
" event.preventDefault();\n",
" return false;\n",
"}\n",
"\n",
"mpl.figure.prototype._key_event_extra = function(event, name) {\n",
" // Handle any extra behaviour associated with a key event\n",
"}\n",
"\n",
"mpl.figure.prototype.key_event = function(event, name) {\n",
"\n",
" // Prevent repeat events\n",
" if (name == 'key_press')\n",
" {\n",
" if (event.which === this._key)\n",
" return;\n",
" else\n",
" this._key = event.which;\n",
" }\n",
" if (name == 'key_release')\n",
" this._key = null;\n",
"\n",
" var value = '';\n",
" if (event.ctrlKey && event.which != 17)\n",
" value += \"ctrl+\";\n",
" if (event.altKey && event.which != 18)\n",
" value += \"alt+\";\n",
" if (event.shiftKey && event.which != 16)\n",
" value += \"shift+\";\n",
"\n",
" value += 'k';\n",
" value += event.which.toString();\n",
"\n",
" this._key_event_extra(event, name);\n",
"\n",
" this.send_message(name, {key: value,\n",
" guiEvent: simpleKeys(event)});\n",
" return false;\n",
"}\n",
"\n",
"mpl.figure.prototype.toolbar_button_onclick = function(name) {\n",
" if (name == 'download') {\n",
" this.handle_save(this, null);\n",
" } else {\n",
" this.send_message(\"toolbar_button\", {name: name});\n",
" }\n",
"};\n",
"\n",
"mpl.figure.prototype.toolbar_button_onmouseover = function(tooltip) {\n",
" this.message.textContent = tooltip;\n",
"};\n",
"mpl.toolbar_items = [[\"Home\", \"Reset original view\", \"fa fa-home icon-home\", \"home\"], [\"Back\", \"Back to previous view\", \"fa fa-arrow-left icon-arrow-left\", \"back\"], [\"Forward\", \"Forward to next view\", \"fa fa-arrow-right icon-arrow-right\", \"forward\"], [\"\", \"\", \"\", \"\"], [\"Pan\", \"Pan axes with left mouse, zoom with right\", \"fa fa-arrows icon-move\", \"pan\"], [\"Zoom\", \"Zoom to rectangle\", \"fa fa-square-o icon-check-empty\", \"zoom\"], [\"\", \"\", \"\", \"\"], [\"Download\", \"Download plot\", \"fa fa-floppy-o icon-save\", \"download\"]];\n",
"\n",
"mpl.extensions = [\"eps\", \"jpeg\", \"pdf\", \"png\", \"ps\", \"raw\", \"svg\", \"tif\"];\n",
"\n",
"mpl.default_extension = \"png\";var comm_websocket_adapter = function(comm) {\n",
" // Create a \"websocket\"-like object which calls the given IPython comm\n",
" // object with the appropriate methods. Currently this is a non binary\n",
" // socket, so there is still some room for performance tuning.\n",
" var ws = {};\n",
"\n",
" ws.close = function() {\n",
" comm.close()\n",
" };\n",
" ws.send = function(m) {\n",
" //console.log('sending', m);\n",
" comm.send(m);\n",
" };\n",
" // Register the callback with on_msg.\n",
" comm.on_msg(function(msg) {\n",
" //console.log('receiving', msg['content']['data'], msg);\n",
" // Pass the mpl event to the overridden (by mpl) onmessage function.\n",
" ws.onmessage(msg['content']['data'])\n",
" });\n",
" return ws;\n",
"}\n",
"\n",
"mpl.mpl_figure_comm = function(comm, msg) {\n",
" // This is the function which gets called when the mpl process\n",
" // starts-up an IPython Comm through the \"matplotlib\" channel.\n",
"\n",
" var id = msg.content.data.id;\n",
" // Get hold of the div created by the display call when the Comm\n",
" // socket was opened in Python.\n",
" var element = $(\"#\" + id);\n",
" var ws_proxy = comm_websocket_adapter(comm)\n",
"\n",
" function ondownload(figure, format) {\n",
" window.open(figure.imageObj.src);\n",
" }\n",
"\n",
" var fig = new mpl.figure(id, ws_proxy,\n",
" ondownload,\n",
" element.get(0));\n",
"\n",
" // Call onopen now - mpl needs it, as it is assuming we've passed it a real\n",
" // web socket which is closed, not our websocket->open comm proxy.\n",
" ws_proxy.onopen();\n",
"\n",
" fig.parent_element = element.get(0);\n",
" fig.cell_info = mpl.find_output_cell(\"\");\n",
" if (!fig.cell_info) {\n",
" console.error(\"Failed to find cell for figure\", id, fig);\n",
" return;\n",
" }\n",
"\n",
" var output_index = fig.cell_info[2]\n",
" var cell = fig.cell_info[0];\n",
"\n",
"};\n",
"\n",
"mpl.figure.prototype.handle_close = function(fig, msg) {\n",
" var width = fig.canvas.width/mpl.ratio\n",
" fig.root.unbind('remove')\n",
"\n",
" // Update the output cell to use the data from the current canvas.\n",
" fig.push_to_output();\n",
" var dataURL = fig.canvas.toDataURL();\n",
" // Re-enable the keyboard manager in IPython - without this line, in FF,\n",
" // the notebook keyboard shortcuts fail.\n",
" IPython.keyboard_manager.enable()\n",
" $(fig.parent_element).html('');\n",
" fig.close_ws(fig, msg);\n",
"}\n",
"\n",
"mpl.figure.prototype.close_ws = function(fig, msg){\n",
" fig.send_message('closing', msg);\n",
" // fig.ws.close()\n",
"}\n",
"\n",
"mpl.figure.prototype.push_to_output = function(remove_interactive) {\n",
" // Turn the data on the canvas into data in the output cell.\n",
" var width = this.canvas.width/mpl.ratio\n",
" var dataURL = this.canvas.toDataURL();\n",
" this.cell_info[1]['text/html'] = '';\n",
"}\n",
"\n",
"mpl.figure.prototype.updated_canvas_event = function() {\n",
" // Tell IPython that the notebook contents must change.\n",
" IPython.notebook.set_dirty(true);\n",
" this.send_message(\"ack\", {});\n",
" var fig = this;\n",
" // Wait a second, then push the new image to the DOM so\n",
" // that it is saved nicely (might be nice to debounce this).\n",
" setTimeout(function () { fig.push_to_output() }, 1000);\n",
"}\n",
"\n",
"mpl.figure.prototype._init_toolbar = function() {\n",
" var fig = this;\n",
"\n",
" var nav_element = $('')\n",
" nav_element.attr('style', 'width: 100%');\n",
" this.root.append(nav_element);\n",
"\n",
" // Define a callback function for later on.\n",
" function toolbar_event(event) {\n",
" return fig.toolbar_button_onclick(event['data']);\n",
" }\n",
" function toolbar_mouse_event(event) {\n",
" return fig.toolbar_button_onmouseover(event['data']);\n",
" }\n",
"\n",
" for(var toolbar_ind in mpl.toolbar_items){\n",
" var name = mpl.toolbar_items[toolbar_ind][0];\n",
" var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
" var image = mpl.toolbar_items[toolbar_ind][2];\n",
" var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
"\n",
" if (!name) { continue; };\n",
"\n",
" var button = $('');\n",
" button.click(method_name, toolbar_event);\n",
" button.mouseover(tooltip, toolbar_mouse_event);\n",
" nav_element.append(button);\n",
" }\n",
"\n",
" // Add the status bar.\n",
" var status_bar = $('');\n",
" nav_element.append(status_bar);\n",
" this.message = status_bar[0];\n",
"\n",
" // Add the close button to the window.\n",
" var buttongrp = $('');\n",
" var button = $('');\n",
" button.click(function (evt) { fig.handle_close(fig, {}); } );\n",
" button.mouseover('Stop Interaction', toolbar_mouse_event);\n",
" buttongrp.append(button);\n",
" var titlebar = this.root.find($('.ui-dialog-titlebar'));\n",
" titlebar.prepend(buttongrp);\n",
"}\n",
"\n",
"mpl.figure.prototype._root_extra_style = function(el){\n",
" var fig = this\n",
" el.on(\"remove\", function(){\n",
"\tfig.close_ws(fig, {});\n",
" });\n",
"}\n",
"\n",
"mpl.figure.prototype._canvas_extra_style = function(el){\n",
" // this is important to make the div 'focusable\n",
" el.attr('tabindex', 0)\n",
" // reach out to IPython and tell the keyboard manager to turn it's self\n",
" // off when our div gets focus\n",
"\n",
" // location in version 3\n",
" if (IPython.notebook.keyboard_manager) {\n",
" IPython.notebook.keyboard_manager.register_events(el);\n",
" }\n",
" else {\n",
" // location in version 2\n",
" IPython.keyboard_manager.register_events(el);\n",
" }\n",
"\n",
"}\n",
"\n",
"mpl.figure.prototype._key_event_extra = function(event, name) {\n",
" var manager = IPython.notebook.keyboard_manager;\n",
" if (!manager)\n",
" manager = IPython.keyboard_manager;\n",
"\n",
" // Check for shift+enter\n",
" if (event.shiftKey && event.which == 13) {\n",
" this.canvas_div.blur();\n",
" event.shiftKey = false;\n",
" // Send a \"J\" for go to next cell\n",
" event.which = 74;\n",
" event.keyCode = 74;\n",
" manager.command_mode();\n",
" manager.handle_keydown(event);\n",
" }\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_save = function(fig, msg) {\n",
" fig.ondownload(fig, null);\n",
"}\n",
"\n",
"\n",
"mpl.find_output_cell = function(html_output) {\n",
" // Return the cell and output element which can be found *uniquely* in the notebook.\n",
" // Note - this is a bit hacky, but it is done because the \"notebook_saving.Notebook\"\n",
" // IPython event is triggered only after the cells have been serialised, which for\n",
" // our purposes (turning an active figure into a static one), is too late.\n",
" var cells = IPython.notebook.get_cells();\n",
" var ncells = cells.length;\n",
" for (var i=0; i= 3 moved mimebundle to data attribute of output\n",
" data = data.data;\n",
" }\n",
" if (data['text/html'] == html_output) {\n",
" return [cell, data, j];\n",
" }\n",
" }\n",
" }\n",
" }\n",
"}\n",
"\n",
"// Register the function which deals with the matplotlib target/channel.\n",
"// The kernel may be null if the page has been refreshed.\n",
"if (IPython.notebook.kernel != null) {\n",
" IPython.notebook.kernel.comm_manager.register_target('matplotlib', mpl.mpl_figure_comm);\n",
"}\n"
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/html": [
""
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/javascript": [
"/* Put everything inside the global mpl namespace */\n",
"window.mpl = {};\n",
"\n",
"\n",
"mpl.get_websocket_type = function() {\n",
" if (typeof(WebSocket) !== 'undefined') {\n",
" return WebSocket;\n",
" } else if (typeof(MozWebSocket) !== 'undefined') {\n",
" return MozWebSocket;\n",
" } else {\n",
" alert('Your browser does not have WebSocket support.' +\n",
" 'Please try Chrome, Safari or Firefox ≥ 6. ' +\n",
" 'Firefox 4 and 5 are also supported but you ' +\n",
" 'have to enable WebSockets in about:config.');\n",
" };\n",
"}\n",
"\n",
"mpl.figure = function(figure_id, websocket, ondownload, parent_element) {\n",
" this.id = figure_id;\n",
"\n",
" this.ws = websocket;\n",
"\n",
" this.supports_binary = (this.ws.binaryType != undefined);\n",
"\n",
" if (!this.supports_binary) {\n",
" var warnings = document.getElementById(\"mpl-warnings\");\n",
" if (warnings) {\n",
" warnings.style.display = 'block';\n",
" warnings.textContent = (\n",
" \"This browser does not support binary websocket messages. \" +\n",
" \"Performance may be slow.\");\n",
" }\n",
" }\n",
"\n",
" this.imageObj = new Image();\n",
"\n",
" this.context = undefined;\n",
" this.message = undefined;\n",
" this.canvas = undefined;\n",
" this.rubberband_canvas = undefined;\n",
" this.rubberband_context = undefined;\n",
" this.format_dropdown = undefined;\n",
"\n",
" this.image_mode = 'full';\n",
"\n",
" this.root = $('');\n",
" this._root_extra_style(this.root)\n",
" this.root.attr('style', 'display: inline-block');\n",
"\n",
" $(parent_element).append(this.root);\n",
"\n",
" this._init_header(this);\n",
" this._init_canvas(this);\n",
" this._init_toolbar(this);\n",
"\n",
" var fig = this;\n",
"\n",
" this.waiting = false;\n",
"\n",
" this.ws.onopen = function () {\n",
" fig.send_message(\"supports_binary\", {value: fig.supports_binary});\n",
" fig.send_message(\"send_image_mode\", {});\n",
" if (mpl.ratio != 1) {\n",
" fig.send_message(\"set_dpi_ratio\", {'dpi_ratio': mpl.ratio});\n",
" }\n",
" fig.send_message(\"refresh\", {});\n",
" }\n",
"\n",
" this.imageObj.onload = function() {\n",
" if (fig.image_mode == 'full') {\n",
" // Full images could contain transparency (where diff images\n",
" // almost always do), so we need to clear the canvas so that\n",
" // there is no ghosting.\n",
" fig.context.clearRect(0, 0, fig.canvas.width, fig.canvas.height);\n",
" }\n",
" fig.context.drawImage(fig.imageObj, 0, 0);\n",
" };\n",
"\n",
" this.imageObj.onunload = function() {\n",
" fig.ws.close();\n",
" }\n",
"\n",
" this.ws.onmessage = this._make_on_message_function(this);\n",
"\n",
" this.ondownload = ondownload;\n",
"}\n",
"\n",
"mpl.figure.prototype._init_header = function() {\n",
" var titlebar = $(\n",
" '');\n",
" var titletext = $(\n",
" '');\n",
" titlebar.append(titletext)\n",
" this.root.append(titlebar);\n",
" this.header = titletext[0];\n",
"}\n",
"\n",
"\n",
"\n",
"mpl.figure.prototype._canvas_extra_style = function(canvas_div) {\n",
"\n",
"}\n",
"\n",
"\n",
"mpl.figure.prototype._root_extra_style = function(canvas_div) {\n",
"\n",
"}\n",
"\n",
"mpl.figure.prototype._init_canvas = function() {\n",
" var fig = this;\n",
"\n",
" var canvas_div = $('');\n",
"\n",
" canvas_div.attr('style', 'position: relative; clear: both; outline: 0');\n",
"\n",
" function canvas_keyboard_event(event) {\n",
" return fig.key_event(event, event['data']);\n",
" }\n",
"\n",
" canvas_div.keydown('key_press', canvas_keyboard_event);\n",
" canvas_div.keyup('key_release', canvas_keyboard_event);\n",
" this.canvas_div = canvas_div\n",
" this._canvas_extra_style(canvas_div)\n",
" this.root.append(canvas_div);\n",
"\n",
" var canvas = $('');\n",
" canvas.addClass('mpl-canvas');\n",
" canvas.attr('style', \"left: 0; top: 0; z-index: 0; outline: 0\")\n",
"\n",
" this.canvas = canvas[0];\n",
" this.context = canvas[0].getContext(\"2d\");\n",
"\n",
" var backingStore = this.context.backingStorePixelRatio ||\n",
"\tthis.context.webkitBackingStorePixelRatio ||\n",
"\tthis.context.mozBackingStorePixelRatio ||\n",
"\tthis.context.msBackingStorePixelRatio ||\n",
"\tthis.context.oBackingStorePixelRatio ||\n",
"\tthis.context.backingStorePixelRatio || 1;\n",
"\n",
" mpl.ratio = (window.devicePixelRatio || 1) / backingStore;\n",
"\n",
" var rubberband = $('');\n",
" rubberband.attr('style', \"position: absolute; left: 0; top: 0; z-index: 1;\")\n",
"\n",
" var pass_mouse_events = true;\n",
"\n",
" canvas_div.resizable({\n",
" start: function(event, ui) {\n",
" pass_mouse_events = false;\n",
" },\n",
" resize: function(event, ui) {\n",
" fig.request_resize(ui.size.width, ui.size.height);\n",
" },\n",
" stop: function(event, ui) {\n",
" pass_mouse_events = true;\n",
" fig.request_resize(ui.size.width, ui.size.height);\n",
" },\n",
" });\n",
"\n",
" function mouse_event_fn(event) {\n",
" if (pass_mouse_events)\n",
" return fig.mouse_event(event, event['data']);\n",
" }\n",
"\n",
" rubberband.mousedown('button_press', mouse_event_fn);\n",
" rubberband.mouseup('button_release', mouse_event_fn);\n",
" // Throttle sequential mouse events to 1 every 20ms.\n",
" rubberband.mousemove('motion_notify', mouse_event_fn);\n",
"\n",
" rubberband.mouseenter('figure_enter', mouse_event_fn);\n",
" rubberband.mouseleave('figure_leave', mouse_event_fn);\n",
"\n",
" canvas_div.on(\"wheel\", function (event) {\n",
" event = event.originalEvent;\n",
" event['data'] = 'scroll'\n",
" if (event.deltaY < 0) {\n",
" event.step = 1;\n",
" } else {\n",
" event.step = -1;\n",
" }\n",
" mouse_event_fn(event);\n",
" });\n",
"\n",
" canvas_div.append(canvas);\n",
" canvas_div.append(rubberband);\n",
"\n",
" this.rubberband = rubberband;\n",
" this.rubberband_canvas = rubberband[0];\n",
" this.rubberband_context = rubberband[0].getContext(\"2d\");\n",
" this.rubberband_context.strokeStyle = \"#000000\";\n",
"\n",
" this._resize_canvas = function(width, height) {\n",
" // Keep the size of the canvas, canvas container, and rubber band\n",
" // canvas in synch.\n",
" canvas_div.css('width', width)\n",
" canvas_div.css('height', height)\n",
"\n",
" canvas.attr('width', width * mpl.ratio);\n",
" canvas.attr('height', height * mpl.ratio);\n",
" canvas.attr('style', 'width: ' + width + 'px; height: ' + height + 'px;');\n",
"\n",
" rubberband.attr('width', width);\n",
" rubberband.attr('height', height);\n",
" }\n",
"\n",
" // Set the figure to an initial 600x600px, this will subsequently be updated\n",
" // upon first draw.\n",
" this._resize_canvas(600, 600);\n",
"\n",
" // Disable right mouse context menu.\n",
" $(this.rubberband_canvas).bind(\"contextmenu\",function(e){\n",
" return false;\n",
" });\n",
"\n",
" function set_focus () {\n",
" canvas.focus();\n",
" canvas_div.focus();\n",
" }\n",
"\n",
" window.setTimeout(set_focus, 100);\n",
"}\n",
"\n",
"mpl.figure.prototype._init_toolbar = function() {\n",
" var fig = this;\n",
"\n",
" var nav_element = $('')\n",
" nav_element.attr('style', 'width: 100%');\n",
" this.root.append(nav_element);\n",
"\n",
" // Define a callback function for later on.\n",
" function toolbar_event(event) {\n",
" return fig.toolbar_button_onclick(event['data']);\n",
" }\n",
" function toolbar_mouse_event(event) {\n",
" return fig.toolbar_button_onmouseover(event['data']);\n",
" }\n",
"\n",
" for(var toolbar_ind in mpl.toolbar_items) {\n",
" var name = mpl.toolbar_items[toolbar_ind][0];\n",
" var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
" var image = mpl.toolbar_items[toolbar_ind][2];\n",
" var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
"\n",
" if (!name) {\n",
" // put a spacer in here.\n",
" continue;\n",
" }\n",
" var button = $('');\n",
" button.addClass('ui-button ui-widget ui-state-default ui-corner-all ' +\n",
" 'ui-button-icon-only');\n",
" button.attr('role', 'button');\n",
" button.attr('aria-disabled', 'false');\n",
" button.click(method_name, toolbar_event);\n",
" button.mouseover(tooltip, toolbar_mouse_event);\n",
"\n",
" var icon_img = $('');\n",
" icon_img.addClass('ui-button-icon-primary ui-icon');\n",
" icon_img.addClass(image);\n",
" icon_img.addClass('ui-corner-all');\n",
"\n",
" var tooltip_span = $('');\n",
" tooltip_span.addClass('ui-button-text');\n",
" tooltip_span.html(tooltip);\n",
"\n",
" button.append(icon_img);\n",
" button.append(tooltip_span);\n",
"\n",
" nav_element.append(button);\n",
" }\n",
"\n",
" var fmt_picker_span = $('');\n",
"\n",
" var fmt_picker = $('');\n",
" fmt_picker.addClass('mpl-toolbar-option ui-widget ui-widget-content');\n",
" fmt_picker_span.append(fmt_picker);\n",
" nav_element.append(fmt_picker_span);\n",
" this.format_dropdown = fmt_picker[0];\n",
"\n",
" for (var ind in mpl.extensions) {\n",
" var fmt = mpl.extensions[ind];\n",
" var option = $(\n",
" '', {selected: fmt === mpl.default_extension}).html(fmt);\n",
" fmt_picker.append(option)\n",
" }\n",
"\n",
" // Add hover states to the ui-buttons\n",
" $( \".ui-button\" ).hover(\n",
" function() { $(this).addClass(\"ui-state-hover\");},\n",
" function() { $(this).removeClass(\"ui-state-hover\");}\n",
" );\n",
"\n",
" var status_bar = $('');\n",
" nav_element.append(status_bar);\n",
" this.message = status_bar[0];\n",
"}\n",
"\n",
"mpl.figure.prototype.request_resize = function(x_pixels, y_pixels) {\n",
" // Request matplotlib to resize the figure. Matplotlib will then trigger a resize in the client,\n",
" // which will in turn request a refresh of the image.\n",
" this.send_message('resize', {'width': x_pixels, 'height': y_pixels});\n",
"}\n",
"\n",
"mpl.figure.prototype.send_message = function(type, properties) {\n",
" properties['type'] = type;\n",
" properties['figure_id'] = this.id;\n",
" this.ws.send(JSON.stringify(properties));\n",
"}\n",
"\n",
"mpl.figure.prototype.send_draw_message = function() {\n",
" if (!this.waiting) {\n",
" this.waiting = true;\n",
" this.ws.send(JSON.stringify({type: \"draw\", figure_id: this.id}));\n",
" }\n",
"}\n",
"\n",
"\n",
"mpl.figure.prototype.handle_save = function(fig, msg) {\n",
" var format_dropdown = fig.format_dropdown;\n",
" var format = format_dropdown.options[format_dropdown.selectedIndex].value;\n",
" fig.ondownload(fig, format);\n",
"}\n",
"\n",
"\n",
"mpl.figure.prototype.handle_resize = function(fig, msg) {\n",
" var size = msg['size'];\n",
" if (size[0] != fig.canvas.width || size[1] != fig.canvas.height) {\n",
" fig._resize_canvas(size[0], size[1]);\n",
" fig.send_message(\"refresh\", {});\n",
" };\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_rubberband = function(fig, msg) {\n",
" var x0 = msg['x0'] / mpl.ratio;\n",
" var y0 = (fig.canvas.height - msg['y0']) / mpl.ratio;\n",
" var x1 = msg['x1'] / mpl.ratio;\n",
" var y1 = (fig.canvas.height - msg['y1']) / mpl.ratio;\n",
" x0 = Math.floor(x0) + 0.5;\n",
" y0 = Math.floor(y0) + 0.5;\n",
" x1 = Math.floor(x1) + 0.5;\n",
" y1 = Math.floor(y1) + 0.5;\n",
" var min_x = Math.min(x0, x1);\n",
" var min_y = Math.min(y0, y1);\n",
" var width = Math.abs(x1 - x0);\n",
" var height = Math.abs(y1 - y0);\n",
"\n",
" fig.rubberband_context.clearRect(\n",
" 0, 0, fig.canvas.width, fig.canvas.height);\n",
"\n",
" fig.rubberband_context.strokeRect(min_x, min_y, width, height);\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_figure_label = function(fig, msg) {\n",
" // Updates the figure title.\n",
" fig.header.textContent = msg['label'];\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_cursor = function(fig, msg) {\n",
" var cursor = msg['cursor'];\n",
" switch(cursor)\n",
" {\n",
" case 0:\n",
" cursor = 'pointer';\n",
" break;\n",
" case 1:\n",
" cursor = 'default';\n",
" break;\n",
" case 2:\n",
" cursor = 'crosshair';\n",
" break;\n",
" case 3:\n",
" cursor = 'move';\n",
" break;\n",
" }\n",
" fig.rubberband_canvas.style.cursor = cursor;\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_message = function(fig, msg) {\n",
" fig.message.textContent = msg['message'];\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_draw = function(fig, msg) {\n",
" // Request the server to send over a new figure.\n",
" fig.send_draw_message();\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_image_mode = function(fig, msg) {\n",
" fig.image_mode = msg['mode'];\n",
"}\n",
"\n",
"mpl.figure.prototype.updated_canvas_event = function() {\n",
" // Called whenever the canvas gets updated.\n",
" this.send_message(\"ack\", {});\n",
"}\n",
"\n",
"// A function to construct a web socket function for onmessage handling.\n",
"// Called in the figure constructor.\n",
"mpl.figure.prototype._make_on_message_function = function(fig) {\n",
" return function socket_on_message(evt) {\n",
" if (evt.data instanceof Blob) {\n",
" /* FIXME: We get \"Resource interpreted as Image but\n",
" * transferred with MIME type text/plain:\" errors on\n",
" * Chrome. But how to set the MIME type? It doesn't seem\n",
" * to be part of the websocket stream */\n",
" evt.data.type = \"image/png\";\n",
"\n",
" /* Free the memory for the previous frames */\n",
" if (fig.imageObj.src) {\n",
" (window.URL || window.webkitURL).revokeObjectURL(\n",
" fig.imageObj.src);\n",
" }\n",
"\n",
" fig.imageObj.src = (window.URL || window.webkitURL).createObjectURL(\n",
" evt.data);\n",
" fig.updated_canvas_event();\n",
" fig.waiting = false;\n",
" return;\n",
" }\n",
" else if (typeof evt.data === 'string' && evt.data.slice(0, 21) == \"data:image/png;base64\") {\n",
" fig.imageObj.src = evt.data;\n",
" fig.updated_canvas_event();\n",
" fig.waiting = false;\n",
" return;\n",
" }\n",
"\n",
" var msg = JSON.parse(evt.data);\n",
" var msg_type = msg['type'];\n",
"\n",
" // Call the \"handle_{type}\" callback, which takes\n",
" // the figure and JSON message as its only arguments.\n",
" try {\n",
" var callback = fig[\"handle_\" + msg_type];\n",
" } catch (e) {\n",
" console.log(\"No handler for the '\" + msg_type + \"' message type: \", msg);\n",
" return;\n",
" }\n",
"\n",
" if (callback) {\n",
" try {\n",
" // console.log(\"Handling '\" + msg_type + \"' message: \", msg);\n",
" callback(fig, msg);\n",
" } catch (e) {\n",
" console.log(\"Exception inside the 'handler_\" + msg_type + \"' callback:\", e, e.stack, msg);\n",
" }\n",
" }\n",
" };\n",
"}\n",
"\n",
"// from http://stackoverflow.com/questions/1114465/getting-mouse-location-in-canvas\n",
"mpl.findpos = function(e) {\n",
" //this section is from http://www.quirksmode.org/js/events_properties.html\n",
" var targ;\n",
" if (!e)\n",
" e = window.event;\n",
" if (e.target)\n",
" targ = e.target;\n",
" else if (e.srcElement)\n",
" targ = e.srcElement;\n",
" if (targ.nodeType == 3) // defeat Safari bug\n",
" targ = targ.parentNode;\n",
"\n",
" // jQuery normalizes the pageX and pageY\n",
" // pageX,Y are the mouse positions relative to the document\n",
" // offset() returns the position of the element relative to the document\n",
" var x = e.pageX - $(targ).offset().left;\n",
" var y = e.pageY - $(targ).offset().top;\n",
"\n",
" return {\"x\": x, \"y\": y};\n",
"};\n",
"\n",
"/*\n",
" * return a copy of an object with only non-object keys\n",
" * we need this to avoid circular references\n",
" * http://stackoverflow.com/a/24161582/3208463\n",
" */\n",
"function simpleKeys (original) {\n",
" return Object.keys(original).reduce(function (obj, key) {\n",
" if (typeof original[key] !== 'object')\n",
" obj[key] = original[key]\n",
" return obj;\n",
" }, {});\n",
"}\n",
"\n",
"mpl.figure.prototype.mouse_event = function(event, name) {\n",
" var canvas_pos = mpl.findpos(event)\n",
"\n",
" if (name === 'button_press')\n",
" {\n",
" this.canvas.focus();\n",
" this.canvas_div.focus();\n",
" }\n",
"\n",
" var x = canvas_pos.x * mpl.ratio;\n",
" var y = canvas_pos.y * mpl.ratio;\n",
"\n",
" this.send_message(name, {x: x, y: y, button: event.button,\n",
" step: event.step,\n",
" guiEvent: simpleKeys(event)});\n",
"\n",
" /* This prevents the web browser from automatically changing to\n",
" * the text insertion cursor when the button is pressed. We want\n",
" * to control all of the cursor setting manually through the\n",
" * 'cursor' event from matplotlib */\n",
" event.preventDefault();\n",
" return false;\n",
"}\n",
"\n",
"mpl.figure.prototype._key_event_extra = function(event, name) {\n",
" // Handle any extra behaviour associated with a key event\n",
"}\n",
"\n",
"mpl.figure.prototype.key_event = function(event, name) {\n",
"\n",
" // Prevent repeat events\n",
" if (name == 'key_press')\n",
" {\n",
" if (event.which === this._key)\n",
" return;\n",
" else\n",
" this._key = event.which;\n",
" }\n",
" if (name == 'key_release')\n",
" this._key = null;\n",
"\n",
" var value = '';\n",
" if (event.ctrlKey && event.which != 17)\n",
" value += \"ctrl+\";\n",
" if (event.altKey && event.which != 18)\n",
" value += \"alt+\";\n",
" if (event.shiftKey && event.which != 16)\n",
" value += \"shift+\";\n",
"\n",
" value += 'k';\n",
" value += event.which.toString();\n",
"\n",
" this._key_event_extra(event, name);\n",
"\n",
" this.send_message(name, {key: value,\n",
" guiEvent: simpleKeys(event)});\n",
" return false;\n",
"}\n",
"\n",
"mpl.figure.prototype.toolbar_button_onclick = function(name) {\n",
" if (name == 'download') {\n",
" this.handle_save(this, null);\n",
" } else {\n",
" this.send_message(\"toolbar_button\", {name: name});\n",
" }\n",
"};\n",
"\n",
"mpl.figure.prototype.toolbar_button_onmouseover = function(tooltip) {\n",
" this.message.textContent = tooltip;\n",
"};\n",
"mpl.toolbar_items = [[\"Home\", \"Reset original view\", \"fa fa-home icon-home\", \"home\"], [\"Back\", \"Back to previous view\", \"fa fa-arrow-left icon-arrow-left\", \"back\"], [\"Forward\", \"Forward to next view\", \"fa fa-arrow-right icon-arrow-right\", \"forward\"], [\"\", \"\", \"\", \"\"], [\"Pan\", \"Pan axes with left mouse, zoom with right\", \"fa fa-arrows icon-move\", \"pan\"], [\"Zoom\", \"Zoom to rectangle\", \"fa fa-square-o icon-check-empty\", \"zoom\"], [\"\", \"\", \"\", \"\"], [\"Download\", \"Download plot\", \"fa fa-floppy-o icon-save\", \"download\"]];\n",
"\n",
"mpl.extensions = [\"eps\", \"jpeg\", \"pdf\", \"png\", \"ps\", \"raw\", \"svg\", \"tif\"];\n",
"\n",
"mpl.default_extension = \"png\";var comm_websocket_adapter = function(comm) {\n",
" // Create a \"websocket\"-like object which calls the given IPython comm\n",
" // object with the appropriate methods. Currently this is a non binary\n",
" // socket, so there is still some room for performance tuning.\n",
" var ws = {};\n",
"\n",
" ws.close = function() {\n",
" comm.close()\n",
" };\n",
" ws.send = function(m) {\n",
" //console.log('sending', m);\n",
" comm.send(m);\n",
" };\n",
" // Register the callback with on_msg.\n",
" comm.on_msg(function(msg) {\n",
" //console.log('receiving', msg['content']['data'], msg);\n",
" // Pass the mpl event to the overridden (by mpl) onmessage function.\n",
" ws.onmessage(msg['content']['data'])\n",
" });\n",
" return ws;\n",
"}\n",
"\n",
"mpl.mpl_figure_comm = function(comm, msg) {\n",
" // This is the function which gets called when the mpl process\n",
" // starts-up an IPython Comm through the \"matplotlib\" channel.\n",
"\n",
" var id = msg.content.data.id;\n",
" // Get hold of the div created by the display call when the Comm\n",
" // socket was opened in Python.\n",
" var element = $(\"#\" + id);\n",
" var ws_proxy = comm_websocket_adapter(comm)\n",
"\n",
" function ondownload(figure, format) {\n",
" window.open(figure.imageObj.src);\n",
" }\n",
"\n",
" var fig = new mpl.figure(id, ws_proxy,\n",
" ondownload,\n",
" element.get(0));\n",
"\n",
" // Call onopen now - mpl needs it, as it is assuming we've passed it a real\n",
" // web socket which is closed, not our websocket->open comm proxy.\n",
" ws_proxy.onopen();\n",
"\n",
" fig.parent_element = element.get(0);\n",
" fig.cell_info = mpl.find_output_cell(\"\");\n",
" if (!fig.cell_info) {\n",
" console.error(\"Failed to find cell for figure\", id, fig);\n",
" return;\n",
" }\n",
"\n",
" var output_index = fig.cell_info[2]\n",
" var cell = fig.cell_info[0];\n",
"\n",
"};\n",
"\n",
"mpl.figure.prototype.handle_close = function(fig, msg) {\n",
" var width = fig.canvas.width/mpl.ratio\n",
" fig.root.unbind('remove')\n",
"\n",
" // Update the output cell to use the data from the current canvas.\n",
" fig.push_to_output();\n",
" var dataURL = fig.canvas.toDataURL();\n",
" // Re-enable the keyboard manager in IPython - without this line, in FF,\n",
" // the notebook keyboard shortcuts fail.\n",
" IPython.keyboard_manager.enable()\n",
" $(fig.parent_element).html('');\n",
" fig.close_ws(fig, msg);\n",
"}\n",
"\n",
"mpl.figure.prototype.close_ws = function(fig, msg){\n",
" fig.send_message('closing', msg);\n",
" // fig.ws.close()\n",
"}\n",
"\n",
"mpl.figure.prototype.push_to_output = function(remove_interactive) {\n",
" // Turn the data on the canvas into data in the output cell.\n",
" var width = this.canvas.width/mpl.ratio\n",
" var dataURL = this.canvas.toDataURL();\n",
" this.cell_info[1]['text/html'] = '';\n",
"}\n",
"\n",
"mpl.figure.prototype.updated_canvas_event = function() {\n",
" // Tell IPython that the notebook contents must change.\n",
" IPython.notebook.set_dirty(true);\n",
" this.send_message(\"ack\", {});\n",
" var fig = this;\n",
" // Wait a second, then push the new image to the DOM so\n",
" // that it is saved nicely (might be nice to debounce this).\n",
" setTimeout(function () { fig.push_to_output() }, 1000);\n",
"}\n",
"\n",
"mpl.figure.prototype._init_toolbar = function() {\n",
" var fig = this;\n",
"\n",
" var nav_element = $('')\n",
" nav_element.attr('style', 'width: 100%');\n",
" this.root.append(nav_element);\n",
"\n",
" // Define a callback function for later on.\n",
" function toolbar_event(event) {\n",
" return fig.toolbar_button_onclick(event['data']);\n",
" }\n",
" function toolbar_mouse_event(event) {\n",
" return fig.toolbar_button_onmouseover(event['data']);\n",
" }\n",
"\n",
" for(var toolbar_ind in mpl.toolbar_items){\n",
" var name = mpl.toolbar_items[toolbar_ind][0];\n",
" var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
" var image = mpl.toolbar_items[toolbar_ind][2];\n",
" var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
"\n",
" if (!name) { continue; };\n",
"\n",
" var button = $('');\n",
" button.click(method_name, toolbar_event);\n",
" button.mouseover(tooltip, toolbar_mouse_event);\n",
" nav_element.append(button);\n",
" }\n",
"\n",
" // Add the status bar.\n",
" var status_bar = $('');\n",
" nav_element.append(status_bar);\n",
" this.message = status_bar[0];\n",
"\n",
" // Add the close button to the window.\n",
" var buttongrp = $('');\n",
" var button = $('');\n",
" button.click(function (evt) { fig.handle_close(fig, {}); } );\n",
" button.mouseover('Stop Interaction', toolbar_mouse_event);\n",
" buttongrp.append(button);\n",
" var titlebar = this.root.find($('.ui-dialog-titlebar'));\n",
" titlebar.prepend(buttongrp);\n",
"}\n",
"\n",
"mpl.figure.prototype._root_extra_style = function(el){\n",
" var fig = this\n",
" el.on(\"remove\", function(){\n",
"\tfig.close_ws(fig, {});\n",
" });\n",
"}\n",
"\n",
"mpl.figure.prototype._canvas_extra_style = function(el){\n",
" // this is important to make the div 'focusable\n",
" el.attr('tabindex', 0)\n",
" // reach out to IPython and tell the keyboard manager to turn it's self\n",
" // off when our div gets focus\n",
"\n",
" // location in version 3\n",
" if (IPython.notebook.keyboard_manager) {\n",
" IPython.notebook.keyboard_manager.register_events(el);\n",
" }\n",
" else {\n",
" // location in version 2\n",
" IPython.keyboard_manager.register_events(el);\n",
" }\n",
"\n",
"}\n",
"\n",
"mpl.figure.prototype._key_event_extra = function(event, name) {\n",
" var manager = IPython.notebook.keyboard_manager;\n",
" if (!manager)\n",
" manager = IPython.keyboard_manager;\n",
"\n",
" // Check for shift+enter\n",
" if (event.shiftKey && event.which == 13) {\n",
" this.canvas_div.blur();\n",
" event.shiftKey = false;\n",
" // Send a \"J\" for go to next cell\n",
" event.which = 74;\n",
" event.keyCode = 74;\n",
" manager.command_mode();\n",
" manager.handle_keydown(event);\n",
" }\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_save = function(fig, msg) {\n",
" fig.ondownload(fig, null);\n",
"}\n",
"\n",
"\n",
"mpl.find_output_cell = function(html_output) {\n",
" // Return the cell and output element which can be found *uniquely* in the notebook.\n",
" // Note - this is a bit hacky, but it is done because the \"notebook_saving.Notebook\"\n",
" // IPython event is triggered only after the cells have been serialised, which for\n",
" // our purposes (turning an active figure into a static one), is too late.\n",
" var cells = IPython.notebook.get_cells();\n",
" var ncells = cells.length;\n",
" for (var i=0; i= 3 moved mimebundle to data attribute of output\n",
" data = data.data;\n",
" }\n",
" if (data['text/html'] == html_output) {\n",
" return [cell, data, j];\n",
" }\n",
" }\n",
" }\n",
" }\n",
"}\n",
"\n",
"// Register the function which deals with the matplotlib target/channel.\n",
"// The kernel may be null if the page has been refreshed.\n",
"if (IPython.notebook.kernel != null) {\n",
" IPython.notebook.kernel.comm_manager.register_target('matplotlib', mpl.mpl_figure_comm);\n",
"}\n"
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/html": [
""
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"#cA3, (cH3, cV3, cD3), (cH2, cV2, cD2), (cH1, cV1, cD1) = coeffs\n",
"J = 3\n",
"coeffs_unchanged = pywt.wavedec2(im_noise, wave, axes=(0,1), level=J)\n",
"shrunk_coeffs = shrink_coeffs(coeffs_unchanged, T)\n",
"\n",
"def plot_hists(c1, c2, ax, ttl):\n",
" bins = np.linspace(-2*c1.std(),2*c1.std(),50)\n",
" ax[0].hist(c1.ravel(), bins=bins, color='r')\n",
" ax[0].set_title(ttl)\n",
" ylim = ax[0].get_ylim()\n",
" \n",
" keep_coeffs = np.abs(c2)>0\n",
" ax[1].hist(c2[keep_coeffs].ravel(), bins=bins)\n",
" ax[1].set_ylim(ylim)\n",
" return keep_coeffs.sum()/keep_coeffs.size\n",
" \n",
"for j in range(J):\n",
" fig, ax = plt.subplots(3,2, figsize=(8,6), gridspec_kw={'hspace': .4})\n",
" fig.suptitle('Scale {}'.format(J-j))\n",
" for b, band in enumerate(['LH', 'HL', 'HH']):\n",
" keep_percent = plot_hists(coeffs_unchanged[1+j][b][...,0], \n",
" shrunk_coeffs[1+j][b][...,0], ax[b], \n",
" ttl='{}, thresh= {:.1f}'.format(band, T[j,b,0]))\n",
" ax[b,1].set_title('{:.1f}% coeffs are zero'.format(100*(1-keep_percent)))"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"We notice that the shrinking has a much larger impact on the finer coefficients. This is where it is easy to gain improvements in SNR as very little signal is here and a lot of noise."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Test commutativity\n",
"Because the inverse wavelet transform is linear, $F(a+b) = F(a) + F(b)$. \n",
"\n",
"MSE is obviously not linear though, so \n",
"$$L(a+b) = \\frac{1}{NM}\\sum_{i,j}(x_{i,j} - \\hat{x}_{i,j})^2 = \\frac{1}{NM}\\sum_{i,j}(x_{i,j} - F(a+b))^2 \\\\\n",
"= \\frac{1}{NM}\\sum_{i,j}\\left(x_{i,j} - (F(a)+F(b))\\right)^2 \\\\\n",
"= L(a) + L(b) - \\frac{1}{NM}\\sum_{i,j}\\left(x_{i,j}^2 - 2F(a)F(b)\\right)$$\n",
"\n",
"The nice thing about orthogonal wavelets is that $\\sum_{i,j}F(a)F(b) = 0$, this means that $L(a+b) = L(a) + L(b) + C$. Then, we can independently minimize each subband's threshold."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Test linearity with subband coefficients\n",
"Annoyingly, we can only use haar for the moment. \n",
"If we us another orthogonal wavelet, the subands are a bigger shape, and \n",
"the above orthogonal property only holds when you set the right wavelet \n",
"coefficients to nonzero."
]
},
{
"cell_type": "code",
"execution_count": 7,
"metadata": {
"ExecuteTime": {
"end_time": "2018-12-03T16:51:24.562470Z",
"start_time": "2018-12-03T16:51:23.836069Z"
}
},
"outputs": [],
"source": [
"coeffs1 = pywt.wavedec2(im_noise, wave, axes=(0,1), level=J)\n",
"coeffs_zero = pywt.wavedec2(np.zeros_like(im_noise), wave, axes=(0,1), level=J)\n",
"# Set level 1 LH to be nonzero\n",
"coeffs_zero[1][0][:] = coeffs1[1][0]\n",
"ya = pywt.waverec2(coeffs_zero, wave, axes=(0,1))\n",
"# Set level 1 HL to also be nonzero\n",
"coeffs_zero[1][1][:] = coeffs1[1][1]\n",
"yab = pywt.waverec2(coeffs_zero, wave, axes=(0,1))\n",
"# Set level 1 LH to be nonzero\n",
"coeffs_zero[1][0][:] = np.zeros_like(coeffs1[1][0])\n",
"yb = pywt.waverec2(coeffs_zero, wave, axes=(0,1))\n",
"np.testing.assert_array_almost_equal(ya+yb, yab, decimal=5)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Test linearity with shrunk subband coefficients"
]
},
{
"cell_type": "code",
"execution_count": 8,
"metadata": {
"ExecuteTime": {
"end_time": "2018-12-03T16:51:25.262910Z",
"start_time": "2018-12-03T16:51:24.564160Z"
}
},
"outputs": [],
"source": [
"coeffs1 = pywt.wavedec2(im_noise, wave, axes=(0,1), level=J)\n",
"coeffs_zero = pywt.wavedec2(np.zeros_like(im_noise), wave, axes=(0,1), level=J)\n",
"\n",
"# Set level 1 LH to be nonzero\n",
"coeffs_zero[1][0][:,:,0] = shrink(coeffs1[1][0][:,:,0], T[0,0,0])\n",
"ya = pywt.waverec2(coeffs_zero, wave, axes=(0,1))\n",
"\n",
"# Set level 1 HL to also be nonzero\n",
"coeffs_zero[1][1][:,:,0] = shrink(coeffs1[1][1][:,:,0], T[0,1,0])\n",
"yab = pywt.waverec2(coeffs_zero, wave, axes=(0,1))\n",
"\n",
"# Set level 1 LH to be nonzero\n",
"coeffs_zero[1][0][:,:,0] = np.zeros_like(coeffs1[1][0][:,:,0])\n",
"yb = pywt.waverec2(coeffs_zero, wave, axes=(0,1))\n",
"np.testing.assert_array_almost_equal(ya+yb, yab, decimal=5)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Test Orthogonality\n",
"Above I have stated that $\\sum_{i,j}F(a)F(b) = 0$. Let us test this"
]
},
{
"cell_type": "code",
"execution_count": 9,
"metadata": {
"ExecuteTime": {
"end_time": "2018-12-03T16:51:27.490477Z",
"start_time": "2018-12-03T16:51:25.264681Z"
}
},
"outputs": [],
"source": [
"coeffs_zero = pywt.wavedec2(np.zeros_like(im_noise), wave, axes=(0,1), level=J)\n",
"coeffs_zero[0][:] = np.random.randn(*coeffs_zero[0].shape)\n",
"ya = pywt.waverec2(coeffs_zero, wave, axes=(0,1))\n",
"\n",
"coeffs_zero = pywt.wavedec2(np.zeros_like(im_noise), wave, axes=(0,1), level=J)\n",
"coeffs_zero[1][1][:] = np.random.randn(*coeffs_zero[1][0].shape)\n",
"yb = pywt.waverec2(coeffs_zero, wave, axes=(0,1))\n",
"\n",
"coeffs_zero = pywt.wavedec2(np.zeros_like(im_noise), wave, axes=(0,1), level=J)\n",
"coeffs_zero[1][2][:] = np.random.randn(*coeffs_zero[1][0].shape)\n",
"yc = pywt.waverec2(coeffs_zero, wave, axes=(0,1))\n",
"\n",
"coeffs_zero = pywt.wavedec2(np.zeros_like(im_noise), wave, axes=(0,1), level=J)\n",
"coeffs_zero[2][0][:] = np.random.randn(*coeffs_zero[2][0].shape)\n",
"yd = pywt.waverec2(coeffs_zero, wave, axes=(0,1))\n",
"\n",
"coeffs_zero = pywt.wavedec2(np.zeros_like(im_noise), wave, axes=(0,1), level=J)\n",
"coeffs_zero[3][2][:] = np.random.randn(*coeffs_zero[3][2].shape)\n",
"ye = pywt.waverec2(coeffs_zero, wave, axes=(0,1))\n",
"\n",
"\n",
"np.testing.assert_array_almost_equal(np.sum(ya*yb), np.zeros_like(ya))\n",
"np.testing.assert_array_almost_equal(np.sum(ya*yc), np.zeros_like(ya))\n",
"np.testing.assert_array_almost_equal(np.sum(ya*yd), np.zeros_like(ya))\n",
"np.testing.assert_array_almost_equal(np.sum(ya*ye), np.zeros_like(ya))\n",
"np.testing.assert_array_almost_equal(np.sum(yb*yc), np.zeros_like(ya))\n",
"np.testing.assert_array_almost_equal(np.sum(yb*yd), np.zeros_like(ya))\n",
"np.testing.assert_array_almost_equal(np.sum(yb*ye), np.zeros_like(ya))\n",
"np.testing.assert_array_almost_equal(np.sum(yc*yd), np.zeros_like(ya))\n",
"np.testing.assert_array_almost_equal(np.sum(yc*ye), np.zeros_like(ya))\n",
"np.testing.assert_array_almost_equal(np.sum(yd*ye), np.zeros_like(ya))"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Test Loss equations"
]
},
{
"cell_type": "code",
"execution_count": 10,
"metadata": {
"ExecuteTime": {
"end_time": "2018-12-03T16:51:28.809385Z",
"start_time": "2018-12-03T16:51:27.491673Z"
}
},
"outputs": [],
"source": [
"def mse_loss(x, x_hat=im):\n",
" return np.mean((x-x_hat)**2)\n",
"\n",
"# Calculate the mean removed image\n",
"coeffs = pywt.wavedec2(im_noise, wave, axes=(0,1), level=J)\n",
"coeffs_zero = pywt.wavedec2(np.zeros_like(im_noise), wave, axes=(0,1), level=J)\n",
"coeffs_zero[0] = np.copy(coeffs[0])\n",
"μ = pywt.waverec2(coeffs_zero, wave, axes=(0,1))\n",
"im2 = im - μ\n",
"\n",
"# let F(a) be passing through level 1 LH\n",
"coeffs_zero = pywt.wavedec2(np.zeros_like(im_noise), wave, axes=(0,1), level=J)\n",
"coeffs_zero[1][0][:,:,0] = shrink(coeffs[1][0][:,:,0], T[0,0,0])\n",
"Fa = pywt.waverec2(coeffs_zero, wave, axes=(0,1))\n",
"La = mse_loss(Fa, im2)\n",
"\n",
"# let F(b) be passing through level 1 HL\n",
"coeffs_zero = pywt.wavedec2(np.zeros_like(im_noise), wave, axes=(0,1), level=J)\n",
"coeffs_zero[1][1][:,:,0] = shrink(coeffs[1][1][:,:,0], T[0,1,0])\n",
"Fb = pywt.waverec2(coeffs_zero, wave, axes=(0,1))\n",
"Lb = mse_loss(Fb, im2)\n",
"\n",
"coeffs_zero = pywt.wavedec2(np.zeros_like(im_noise), wave, axes=(0,1), level=J)\n",
"coeffs_zero[1][0][:,:,0] = shrink(coeffs[1][0][:,:,0], T[0,0,0])\n",
"coeffs_zero[1][1][:,:,0] = shrink(coeffs[1][1][:,:,0], T[0,1,0])\n",
"Fab = pywt.waverec2(coeffs_zero, wave, axes=(0,1))\n",
"Lab = mse_loss(Fab, im2)\n",
"\n",
"np.testing.assert_almost_equal(Lab, La + Lb - np.mean((im2)**2))"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"To expand on the loss equations, consider $L(a,b, \\ldots , z)$. By simple induction, we can say\n",
"$$L(a+b+\\ldots +z) = L(a) + L(b+\\ldots +z) - \\frac{1}{NM}\\sum_{i,j}(x_{i,j})^2$$\n",
"and\n",
"$$L(a+b+\\mu) = L(a+\\mu) + L(b+\\mu) - \\frac{1}{NM}\\sum_{i,j}(x_{i,j} - F(\\mu))^2$$\n",
"Where $F(\\mu)$ is the inverse wavelet transform of the non-zero mean coefficients (in our example, this is simply just the LL coefficients)."
]
},
{
"cell_type": "code",
"execution_count": 11,
"metadata": {
"ExecuteTime": {
"end_time": "2018-12-03T16:51:29.614869Z",
"start_time": "2018-12-03T16:51:28.810874Z"
}
},
"outputs": [],
"source": [
"coeffs = pywt.wavedec2(im_noise, wave, axes=(0,1), level=J)\n",
"coeffs_zero = pywt.wavedec2(np.zeros_like(im_noise), wave, axes=(0,1), level=J)\n",
"\n",
"coeffs_zero[0] = coeffs[0]\n",
"Fμ = pywt.waverec2(coeffs_zero, wave, axes=(0,1))\n",
"\n",
"coeffs_zero[1][0][:,:,0] = shrink(np.copy(coeffs[1][0][:,:,0]), T[0,0,0])\n",
"Faμ = pywt.waverec2(coeffs_zero, wave, axes=(0,1))\n",
"Laμ = mse_loss(Faμ)\n",
"\n",
"coeffs_zero[1][1][:,:,0] = shrink(np.copy(coeffs[1][1][:,:,0]), T[0,1,0])\n",
"Fabμ = pywt.waverec2(coeffs_zero, wave, axes=(0,1))\n",
"Labμ = mse_loss(Fabμ)\n",
"\n",
"coeffs_zero[1][0][:,:,0] = np.zeros_like(coeffs[1][0][:,:,0])\n",
"Fbμ = pywt.waverec2(coeffs_zero, wave, axes=(0,1))\n",
"Lbμ = mse_loss(Fbμ)\n",
"\n",
"np.testing.assert_almost_equal(Labμ, Laμ + Lbμ - np.mean((im-Fμ)**2))"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# Plot MSE as a function of Threshold for each subband\n",
"This nice separability of the loss means we can minimize $L(a)$ wrt $T_a$, then do the same for $L(b)$, and so on. This should give us a global minimum."
]
},
{
"cell_type": "code",
"execution_count": 12,
"metadata": {
"ExecuteTime": {
"end_time": "2018-12-03T16:52:37.414827Z",
"start_time": "2018-12-03T16:51:29.616043Z"
}
},
"outputs": [],
"source": [
"# Calculate the mean removed image\n",
"coeffs = pywt.wavedec2(im_noise, wave, axes=(0,1), level=J)\n",
"coeffs_zero = pywt.wavedec2(np.zeros_like(im_noise), wave, axes=(0,1), level=J)\n",
"coeffs_zero[0] = np.copy(coeffs[0])\n",
"μ = pywt.waverec2(coeffs_zero, wave, axes=(0,1))\n",
"im2 = im - μ\n",
"\n",
"# Iterate over all the subbands, setting all others to zero and find the \n",
"# MSE of the reconstructed images for a range of thresholds.\n",
"N = 20\n",
"J = 3\n",
"C = 3\n",
"mses = np.zeros((J,3,C,N))\n",
"T_mse = np.zeros((J,3,C))\n",
"Ts = np.zeros((J,3,C,N))\n",
"for j in range(J):\n",
" for b, band in enumerate(['LH', 'HL', 'HH']):\n",
" for c in range(C):\n",
" coeffs_zero = pywt.wavedec2(\n",
" np.zeros_like(im_noise), wave, axes=(0,1), level=J)\n",
" k = np.copy(coeffs[1+j][b][:,:,c])\n",
" \n",
" # Calculate 20 points from 0 to double the bayes estimated thresh\n",
" Ts[j,b,c,:] = np.linspace(\n",
" 0, 3*min(T[j,b,c], .4*np.abs(k).max()), N)\n",
" \n",
" # Calculate the mse at the bayes estimate thresh\n",
" coeffs_zero[1+j][b][:,:,c] = shrink(k, T[j,b,c])\n",
" x_hat = pywt.waverec2(coeffs_zero, wave, axes=(0,1))\n",
" T_mse[j,b,c] = np.mean((x_hat - im2)**2)\n",
" \n",
" # Calculate the mse at each of the N points\n",
" for i,t in enumerate(Ts[j,b,c]):\n",
" coeffs_zero[1+j][b][:,:,c] = shrink(k, t)\n",
" x_hat = pywt.waverec2(coeffs_zero, wave, axes=(0,1))\n",
" mses[j,b,c,i] = np.mean((x_hat - im2)**2)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Now that we have calculated the MSEs for a range of thresholds for each subband, plot the results"
]
},
{
"cell_type": "code",
"execution_count": 13,
"metadata": {
"ExecuteTime": {
"end_time": "2018-12-03T16:52:37.819383Z",
"start_time": "2018-12-03T16:52:37.416485Z"
},
"scrolled": false
},
"outputs": [
{
"data": {
"application/javascript": [
"/* Put everything inside the global mpl namespace */\n",
"window.mpl = {};\n",
"\n",
"\n",
"mpl.get_websocket_type = function() {\n",
" if (typeof(WebSocket) !== 'undefined') {\n",
" return WebSocket;\n",
" } else if (typeof(MozWebSocket) !== 'undefined') {\n",
" return MozWebSocket;\n",
" } else {\n",
" alert('Your browser does not have WebSocket support.' +\n",
" 'Please try Chrome, Safari or Firefox ≥ 6. ' +\n",
" 'Firefox 4 and 5 are also supported but you ' +\n",
" 'have to enable WebSockets in about:config.');\n",
" };\n",
"}\n",
"\n",
"mpl.figure = function(figure_id, websocket, ondownload, parent_element) {\n",
" this.id = figure_id;\n",
"\n",
" this.ws = websocket;\n",
"\n",
" this.supports_binary = (this.ws.binaryType != undefined);\n",
"\n",
" if (!this.supports_binary) {\n",
" var warnings = document.getElementById(\"mpl-warnings\");\n",
" if (warnings) {\n",
" warnings.style.display = 'block';\n",
" warnings.textContent = (\n",
" \"This browser does not support binary websocket messages. \" +\n",
" \"Performance may be slow.\");\n",
" }\n",
" }\n",
"\n",
" this.imageObj = new Image();\n",
"\n",
" this.context = undefined;\n",
" this.message = undefined;\n",
" this.canvas = undefined;\n",
" this.rubberband_canvas = undefined;\n",
" this.rubberband_context = undefined;\n",
" this.format_dropdown = undefined;\n",
"\n",
" this.image_mode = 'full';\n",
"\n",
" this.root = $('');\n",
" this._root_extra_style(this.root)\n",
" this.root.attr('style', 'display: inline-block');\n",
"\n",
" $(parent_element).append(this.root);\n",
"\n",
" this._init_header(this);\n",
" this._init_canvas(this);\n",
" this._init_toolbar(this);\n",
"\n",
" var fig = this;\n",
"\n",
" this.waiting = false;\n",
"\n",
" this.ws.onopen = function () {\n",
" fig.send_message(\"supports_binary\", {value: fig.supports_binary});\n",
" fig.send_message(\"send_image_mode\", {});\n",
" if (mpl.ratio != 1) {\n",
" fig.send_message(\"set_dpi_ratio\", {'dpi_ratio': mpl.ratio});\n",
" }\n",
" fig.send_message(\"refresh\", {});\n",
" }\n",
"\n",
" this.imageObj.onload = function() {\n",
" if (fig.image_mode == 'full') {\n",
" // Full images could contain transparency (where diff images\n",
" // almost always do), so we need to clear the canvas so that\n",
" // there is no ghosting.\n",
" fig.context.clearRect(0, 0, fig.canvas.width, fig.canvas.height);\n",
" }\n",
" fig.context.drawImage(fig.imageObj, 0, 0);\n",
" };\n",
"\n",
" this.imageObj.onunload = function() {\n",
" fig.ws.close();\n",
" }\n",
"\n",
" this.ws.onmessage = this._make_on_message_function(this);\n",
"\n",
" this.ondownload = ondownload;\n",
"}\n",
"\n",
"mpl.figure.prototype._init_header = function() {\n",
" var titlebar = $(\n",
" '');\n",
" var titletext = $(\n",
" '');\n",
" titlebar.append(titletext)\n",
" this.root.append(titlebar);\n",
" this.header = titletext[0];\n",
"}\n",
"\n",
"\n",
"\n",
"mpl.figure.prototype._canvas_extra_style = function(canvas_div) {\n",
"\n",
"}\n",
"\n",
"\n",
"mpl.figure.prototype._root_extra_style = function(canvas_div) {\n",
"\n",
"}\n",
"\n",
"mpl.figure.prototype._init_canvas = function() {\n",
" var fig = this;\n",
"\n",
" var canvas_div = $('');\n",
"\n",
" canvas_div.attr('style', 'position: relative; clear: both; outline: 0');\n",
"\n",
" function canvas_keyboard_event(event) {\n",
" return fig.key_event(event, event['data']);\n",
" }\n",
"\n",
" canvas_div.keydown('key_press', canvas_keyboard_event);\n",
" canvas_div.keyup('key_release', canvas_keyboard_event);\n",
" this.canvas_div = canvas_div\n",
" this._canvas_extra_style(canvas_div)\n",
" this.root.append(canvas_div);\n",
"\n",
" var canvas = $('');\n",
" canvas.addClass('mpl-canvas');\n",
" canvas.attr('style', \"left: 0; top: 0; z-index: 0; outline: 0\")\n",
"\n",
" this.canvas = canvas[0];\n",
" this.context = canvas[0].getContext(\"2d\");\n",
"\n",
" var backingStore = this.context.backingStorePixelRatio ||\n",
"\tthis.context.webkitBackingStorePixelRatio ||\n",
"\tthis.context.mozBackingStorePixelRatio ||\n",
"\tthis.context.msBackingStorePixelRatio ||\n",
"\tthis.context.oBackingStorePixelRatio ||\n",
"\tthis.context.backingStorePixelRatio || 1;\n",
"\n",
" mpl.ratio = (window.devicePixelRatio || 1) / backingStore;\n",
"\n",
" var rubberband = $('');\n",
" rubberband.attr('style', \"position: absolute; left: 0; top: 0; z-index: 1;\")\n",
"\n",
" var pass_mouse_events = true;\n",
"\n",
" canvas_div.resizable({\n",
" start: function(event, ui) {\n",
" pass_mouse_events = false;\n",
" },\n",
" resize: function(event, ui) {\n",
" fig.request_resize(ui.size.width, ui.size.height);\n",
" },\n",
" stop: function(event, ui) {\n",
" pass_mouse_events = true;\n",
" fig.request_resize(ui.size.width, ui.size.height);\n",
" },\n",
" });\n",
"\n",
" function mouse_event_fn(event) {\n",
" if (pass_mouse_events)\n",
" return fig.mouse_event(event, event['data']);\n",
" }\n",
"\n",
" rubberband.mousedown('button_press', mouse_event_fn);\n",
" rubberband.mouseup('button_release', mouse_event_fn);\n",
" // Throttle sequential mouse events to 1 every 20ms.\n",
" rubberband.mousemove('motion_notify', mouse_event_fn);\n",
"\n",
" rubberband.mouseenter('figure_enter', mouse_event_fn);\n",
" rubberband.mouseleave('figure_leave', mouse_event_fn);\n",
"\n",
" canvas_div.on(\"wheel\", function (event) {\n",
" event = event.originalEvent;\n",
" event['data'] = 'scroll'\n",
" if (event.deltaY < 0) {\n",
" event.step = 1;\n",
" } else {\n",
" event.step = -1;\n",
" }\n",
" mouse_event_fn(event);\n",
" });\n",
"\n",
" canvas_div.append(canvas);\n",
" canvas_div.append(rubberband);\n",
"\n",
" this.rubberband = rubberband;\n",
" this.rubberband_canvas = rubberband[0];\n",
" this.rubberband_context = rubberband[0].getContext(\"2d\");\n",
" this.rubberband_context.strokeStyle = \"#000000\";\n",
"\n",
" this._resize_canvas = function(width, height) {\n",
" // Keep the size of the canvas, canvas container, and rubber band\n",
" // canvas in synch.\n",
" canvas_div.css('width', width)\n",
" canvas_div.css('height', height)\n",
"\n",
" canvas.attr('width', width * mpl.ratio);\n",
" canvas.attr('height', height * mpl.ratio);\n",
" canvas.attr('style', 'width: ' + width + 'px; height: ' + height + 'px;');\n",
"\n",
" rubberband.attr('width', width);\n",
" rubberband.attr('height', height);\n",
" }\n",
"\n",
" // Set the figure to an initial 600x600px, this will subsequently be updated\n",
" // upon first draw.\n",
" this._resize_canvas(600, 600);\n",
"\n",
" // Disable right mouse context menu.\n",
" $(this.rubberband_canvas).bind(\"contextmenu\",function(e){\n",
" return false;\n",
" });\n",
"\n",
" function set_focus () {\n",
" canvas.focus();\n",
" canvas_div.focus();\n",
" }\n",
"\n",
" window.setTimeout(set_focus, 100);\n",
"}\n",
"\n",
"mpl.figure.prototype._init_toolbar = function() {\n",
" var fig = this;\n",
"\n",
" var nav_element = $('')\n",
" nav_element.attr('style', 'width: 100%');\n",
" this.root.append(nav_element);\n",
"\n",
" // Define a callback function for later on.\n",
" function toolbar_event(event) {\n",
" return fig.toolbar_button_onclick(event['data']);\n",
" }\n",
" function toolbar_mouse_event(event) {\n",
" return fig.toolbar_button_onmouseover(event['data']);\n",
" }\n",
"\n",
" for(var toolbar_ind in mpl.toolbar_items) {\n",
" var name = mpl.toolbar_items[toolbar_ind][0];\n",
" var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
" var image = mpl.toolbar_items[toolbar_ind][2];\n",
" var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
"\n",
" if (!name) {\n",
" // put a spacer in here.\n",
" continue;\n",
" }\n",
" var button = $('');\n",
" button.addClass('ui-button ui-widget ui-state-default ui-corner-all ' +\n",
" 'ui-button-icon-only');\n",
" button.attr('role', 'button');\n",
" button.attr('aria-disabled', 'false');\n",
" button.click(method_name, toolbar_event);\n",
" button.mouseover(tooltip, toolbar_mouse_event);\n",
"\n",
" var icon_img = $('');\n",
" icon_img.addClass('ui-button-icon-primary ui-icon');\n",
" icon_img.addClass(image);\n",
" icon_img.addClass('ui-corner-all');\n",
"\n",
" var tooltip_span = $('');\n",
" tooltip_span.addClass('ui-button-text');\n",
" tooltip_span.html(tooltip);\n",
"\n",
" button.append(icon_img);\n",
" button.append(tooltip_span);\n",
"\n",
" nav_element.append(button);\n",
" }\n",
"\n",
" var fmt_picker_span = $('');\n",
"\n",
" var fmt_picker = $('');\n",
" fmt_picker.addClass('mpl-toolbar-option ui-widget ui-widget-content');\n",
" fmt_picker_span.append(fmt_picker);\n",
" nav_element.append(fmt_picker_span);\n",
" this.format_dropdown = fmt_picker[0];\n",
"\n",
" for (var ind in mpl.extensions) {\n",
" var fmt = mpl.extensions[ind];\n",
" var option = $(\n",
" '', {selected: fmt === mpl.default_extension}).html(fmt);\n",
" fmt_picker.append(option)\n",
" }\n",
"\n",
" // Add hover states to the ui-buttons\n",
" $( \".ui-button\" ).hover(\n",
" function() { $(this).addClass(\"ui-state-hover\");},\n",
" function() { $(this).removeClass(\"ui-state-hover\");}\n",
" );\n",
"\n",
" var status_bar = $('');\n",
" nav_element.append(status_bar);\n",
" this.message = status_bar[0];\n",
"}\n",
"\n",
"mpl.figure.prototype.request_resize = function(x_pixels, y_pixels) {\n",
" // Request matplotlib to resize the figure. Matplotlib will then trigger a resize in the client,\n",
" // which will in turn request a refresh of the image.\n",
" this.send_message('resize', {'width': x_pixels, 'height': y_pixels});\n",
"}\n",
"\n",
"mpl.figure.prototype.send_message = function(type, properties) {\n",
" properties['type'] = type;\n",
" properties['figure_id'] = this.id;\n",
" this.ws.send(JSON.stringify(properties));\n",
"}\n",
"\n",
"mpl.figure.prototype.send_draw_message = function() {\n",
" if (!this.waiting) {\n",
" this.waiting = true;\n",
" this.ws.send(JSON.stringify({type: \"draw\", figure_id: this.id}));\n",
" }\n",
"}\n",
"\n",
"\n",
"mpl.figure.prototype.handle_save = function(fig, msg) {\n",
" var format_dropdown = fig.format_dropdown;\n",
" var format = format_dropdown.options[format_dropdown.selectedIndex].value;\n",
" fig.ondownload(fig, format);\n",
"}\n",
"\n",
"\n",
"mpl.figure.prototype.handle_resize = function(fig, msg) {\n",
" var size = msg['size'];\n",
" if (size[0] != fig.canvas.width || size[1] != fig.canvas.height) {\n",
" fig._resize_canvas(size[0], size[1]);\n",
" fig.send_message(\"refresh\", {});\n",
" };\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_rubberband = function(fig, msg) {\n",
" var x0 = msg['x0'] / mpl.ratio;\n",
" var y0 = (fig.canvas.height - msg['y0']) / mpl.ratio;\n",
" var x1 = msg['x1'] / mpl.ratio;\n",
" var y1 = (fig.canvas.height - msg['y1']) / mpl.ratio;\n",
" x0 = Math.floor(x0) + 0.5;\n",
" y0 = Math.floor(y0) + 0.5;\n",
" x1 = Math.floor(x1) + 0.5;\n",
" y1 = Math.floor(y1) + 0.5;\n",
" var min_x = Math.min(x0, x1);\n",
" var min_y = Math.min(y0, y1);\n",
" var width = Math.abs(x1 - x0);\n",
" var height = Math.abs(y1 - y0);\n",
"\n",
" fig.rubberband_context.clearRect(\n",
" 0, 0, fig.canvas.width, fig.canvas.height);\n",
"\n",
" fig.rubberband_context.strokeRect(min_x, min_y, width, height);\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_figure_label = function(fig, msg) {\n",
" // Updates the figure title.\n",
" fig.header.textContent = msg['label'];\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_cursor = function(fig, msg) {\n",
" var cursor = msg['cursor'];\n",
" switch(cursor)\n",
" {\n",
" case 0:\n",
" cursor = 'pointer';\n",
" break;\n",
" case 1:\n",
" cursor = 'default';\n",
" break;\n",
" case 2:\n",
" cursor = 'crosshair';\n",
" break;\n",
" case 3:\n",
" cursor = 'move';\n",
" break;\n",
" }\n",
" fig.rubberband_canvas.style.cursor = cursor;\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_message = function(fig, msg) {\n",
" fig.message.textContent = msg['message'];\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_draw = function(fig, msg) {\n",
" // Request the server to send over a new figure.\n",
" fig.send_draw_message();\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_image_mode = function(fig, msg) {\n",
" fig.image_mode = msg['mode'];\n",
"}\n",
"\n",
"mpl.figure.prototype.updated_canvas_event = function() {\n",
" // Called whenever the canvas gets updated.\n",
" this.send_message(\"ack\", {});\n",
"}\n",
"\n",
"// A function to construct a web socket function for onmessage handling.\n",
"// Called in the figure constructor.\n",
"mpl.figure.prototype._make_on_message_function = function(fig) {\n",
" return function socket_on_message(evt) {\n",
" if (evt.data instanceof Blob) {\n",
" /* FIXME: We get \"Resource interpreted as Image but\n",
" * transferred with MIME type text/plain:\" errors on\n",
" * Chrome. But how to set the MIME type? It doesn't seem\n",
" * to be part of the websocket stream */\n",
" evt.data.type = \"image/png\";\n",
"\n",
" /* Free the memory for the previous frames */\n",
" if (fig.imageObj.src) {\n",
" (window.URL || window.webkitURL).revokeObjectURL(\n",
" fig.imageObj.src);\n",
" }\n",
"\n",
" fig.imageObj.src = (window.URL || window.webkitURL).createObjectURL(\n",
" evt.data);\n",
" fig.updated_canvas_event();\n",
" fig.waiting = false;\n",
" return;\n",
" }\n",
" else if (typeof evt.data === 'string' && evt.data.slice(0, 21) == \"data:image/png;base64\") {\n",
" fig.imageObj.src = evt.data;\n",
" fig.updated_canvas_event();\n",
" fig.waiting = false;\n",
" return;\n",
" }\n",
"\n",
" var msg = JSON.parse(evt.data);\n",
" var msg_type = msg['type'];\n",
"\n",
" // Call the \"handle_{type}\" callback, which takes\n",
" // the figure and JSON message as its only arguments.\n",
" try {\n",
" var callback = fig[\"handle_\" + msg_type];\n",
" } catch (e) {\n",
" console.log(\"No handler for the '\" + msg_type + \"' message type: \", msg);\n",
" return;\n",
" }\n",
"\n",
" if (callback) {\n",
" try {\n",
" // console.log(\"Handling '\" + msg_type + \"' message: \", msg);\n",
" callback(fig, msg);\n",
" } catch (e) {\n",
" console.log(\"Exception inside the 'handler_\" + msg_type + \"' callback:\", e, e.stack, msg);\n",
" }\n",
" }\n",
" };\n",
"}\n",
"\n",
"// from http://stackoverflow.com/questions/1114465/getting-mouse-location-in-canvas\n",
"mpl.findpos = function(e) {\n",
" //this section is from http://www.quirksmode.org/js/events_properties.html\n",
" var targ;\n",
" if (!e)\n",
" e = window.event;\n",
" if (e.target)\n",
" targ = e.target;\n",
" else if (e.srcElement)\n",
" targ = e.srcElement;\n",
" if (targ.nodeType == 3) // defeat Safari bug\n",
" targ = targ.parentNode;\n",
"\n",
" // jQuery normalizes the pageX and pageY\n",
" // pageX,Y are the mouse positions relative to the document\n",
" // offset() returns the position of the element relative to the document\n",
" var x = e.pageX - $(targ).offset().left;\n",
" var y = e.pageY - $(targ).offset().top;\n",
"\n",
" return {\"x\": x, \"y\": y};\n",
"};\n",
"\n",
"/*\n",
" * return a copy of an object with only non-object keys\n",
" * we need this to avoid circular references\n",
" * http://stackoverflow.com/a/24161582/3208463\n",
" */\n",
"function simpleKeys (original) {\n",
" return Object.keys(original).reduce(function (obj, key) {\n",
" if (typeof original[key] !== 'object')\n",
" obj[key] = original[key]\n",
" return obj;\n",
" }, {});\n",
"}\n",
"\n",
"mpl.figure.prototype.mouse_event = function(event, name) {\n",
" var canvas_pos = mpl.findpos(event)\n",
"\n",
" if (name === 'button_press')\n",
" {\n",
" this.canvas.focus();\n",
" this.canvas_div.focus();\n",
" }\n",
"\n",
" var x = canvas_pos.x * mpl.ratio;\n",
" var y = canvas_pos.y * mpl.ratio;\n",
"\n",
" this.send_message(name, {x: x, y: y, button: event.button,\n",
" step: event.step,\n",
" guiEvent: simpleKeys(event)});\n",
"\n",
" /* This prevents the web browser from automatically changing to\n",
" * the text insertion cursor when the button is pressed. We want\n",
" * to control all of the cursor setting manually through the\n",
" * 'cursor' event from matplotlib */\n",
" event.preventDefault();\n",
" return false;\n",
"}\n",
"\n",
"mpl.figure.prototype._key_event_extra = function(event, name) {\n",
" // Handle any extra behaviour associated with a key event\n",
"}\n",
"\n",
"mpl.figure.prototype.key_event = function(event, name) {\n",
"\n",
" // Prevent repeat events\n",
" if (name == 'key_press')\n",
" {\n",
" if (event.which === this._key)\n",
" return;\n",
" else\n",
" this._key = event.which;\n",
" }\n",
" if (name == 'key_release')\n",
" this._key = null;\n",
"\n",
" var value = '';\n",
" if (event.ctrlKey && event.which != 17)\n",
" value += \"ctrl+\";\n",
" if (event.altKey && event.which != 18)\n",
" value += \"alt+\";\n",
" if (event.shiftKey && event.which != 16)\n",
" value += \"shift+\";\n",
"\n",
" value += 'k';\n",
" value += event.which.toString();\n",
"\n",
" this._key_event_extra(event, name);\n",
"\n",
" this.send_message(name, {key: value,\n",
" guiEvent: simpleKeys(event)});\n",
" return false;\n",
"}\n",
"\n",
"mpl.figure.prototype.toolbar_button_onclick = function(name) {\n",
" if (name == 'download') {\n",
" this.handle_save(this, null);\n",
" } else {\n",
" this.send_message(\"toolbar_button\", {name: name});\n",
" }\n",
"};\n",
"\n",
"mpl.figure.prototype.toolbar_button_onmouseover = function(tooltip) {\n",
" this.message.textContent = tooltip;\n",
"};\n",
"mpl.toolbar_items = [[\"Home\", \"Reset original view\", \"fa fa-home icon-home\", \"home\"], [\"Back\", \"Back to previous view\", \"fa fa-arrow-left icon-arrow-left\", \"back\"], [\"Forward\", \"Forward to next view\", \"fa fa-arrow-right icon-arrow-right\", \"forward\"], [\"\", \"\", \"\", \"\"], [\"Pan\", \"Pan axes with left mouse, zoom with right\", \"fa fa-arrows icon-move\", \"pan\"], [\"Zoom\", \"Zoom to rectangle\", \"fa fa-square-o icon-check-empty\", \"zoom\"], [\"\", \"\", \"\", \"\"], [\"Download\", \"Download plot\", \"fa fa-floppy-o icon-save\", \"download\"]];\n",
"\n",
"mpl.extensions = [\"eps\", \"jpeg\", \"pdf\", \"png\", \"ps\", \"raw\", \"svg\", \"tif\"];\n",
"\n",
"mpl.default_extension = \"png\";var comm_websocket_adapter = function(comm) {\n",
" // Create a \"websocket\"-like object which calls the given IPython comm\n",
" // object with the appropriate methods. Currently this is a non binary\n",
" // socket, so there is still some room for performance tuning.\n",
" var ws = {};\n",
"\n",
" ws.close = function() {\n",
" comm.close()\n",
" };\n",
" ws.send = function(m) {\n",
" //console.log('sending', m);\n",
" comm.send(m);\n",
" };\n",
" // Register the callback with on_msg.\n",
" comm.on_msg(function(msg) {\n",
" //console.log('receiving', msg['content']['data'], msg);\n",
" // Pass the mpl event to the overridden (by mpl) onmessage function.\n",
" ws.onmessage(msg['content']['data'])\n",
" });\n",
" return ws;\n",
"}\n",
"\n",
"mpl.mpl_figure_comm = function(comm, msg) {\n",
" // This is the function which gets called when the mpl process\n",
" // starts-up an IPython Comm through the \"matplotlib\" channel.\n",
"\n",
" var id = msg.content.data.id;\n",
" // Get hold of the div created by the display call when the Comm\n",
" // socket was opened in Python.\n",
" var element = $(\"#\" + id);\n",
" var ws_proxy = comm_websocket_adapter(comm)\n",
"\n",
" function ondownload(figure, format) {\n",
" window.open(figure.imageObj.src);\n",
" }\n",
"\n",
" var fig = new mpl.figure(id, ws_proxy,\n",
" ondownload,\n",
" element.get(0));\n",
"\n",
" // Call onopen now - mpl needs it, as it is assuming we've passed it a real\n",
" // web socket which is closed, not our websocket->open comm proxy.\n",
" ws_proxy.onopen();\n",
"\n",
" fig.parent_element = element.get(0);\n",
" fig.cell_info = mpl.find_output_cell(\"\");\n",
" if (!fig.cell_info) {\n",
" console.error(\"Failed to find cell for figure\", id, fig);\n",
" return;\n",
" }\n",
"\n",
" var output_index = fig.cell_info[2]\n",
" var cell = fig.cell_info[0];\n",
"\n",
"};\n",
"\n",
"mpl.figure.prototype.handle_close = function(fig, msg) {\n",
" var width = fig.canvas.width/mpl.ratio\n",
" fig.root.unbind('remove')\n",
"\n",
" // Update the output cell to use the data from the current canvas.\n",
" fig.push_to_output();\n",
" var dataURL = fig.canvas.toDataURL();\n",
" // Re-enable the keyboard manager in IPython - without this line, in FF,\n",
" // the notebook keyboard shortcuts fail.\n",
" IPython.keyboard_manager.enable()\n",
" $(fig.parent_element).html('');\n",
" fig.close_ws(fig, msg);\n",
"}\n",
"\n",
"mpl.figure.prototype.close_ws = function(fig, msg){\n",
" fig.send_message('closing', msg);\n",
" // fig.ws.close()\n",
"}\n",
"\n",
"mpl.figure.prototype.push_to_output = function(remove_interactive) {\n",
" // Turn the data on the canvas into data in the output cell.\n",
" var width = this.canvas.width/mpl.ratio\n",
" var dataURL = this.canvas.toDataURL();\n",
" this.cell_info[1]['text/html'] = '';\n",
"}\n",
"\n",
"mpl.figure.prototype.updated_canvas_event = function() {\n",
" // Tell IPython that the notebook contents must change.\n",
" IPython.notebook.set_dirty(true);\n",
" this.send_message(\"ack\", {});\n",
" var fig = this;\n",
" // Wait a second, then push the new image to the DOM so\n",
" // that it is saved nicely (might be nice to debounce this).\n",
" setTimeout(function () { fig.push_to_output() }, 1000);\n",
"}\n",
"\n",
"mpl.figure.prototype._init_toolbar = function() {\n",
" var fig = this;\n",
"\n",
" var nav_element = $('')\n",
" nav_element.attr('style', 'width: 100%');\n",
" this.root.append(nav_element);\n",
"\n",
" // Define a callback function for later on.\n",
" function toolbar_event(event) {\n",
" return fig.toolbar_button_onclick(event['data']);\n",
" }\n",
" function toolbar_mouse_event(event) {\n",
" return fig.toolbar_button_onmouseover(event['data']);\n",
" }\n",
"\n",
" for(var toolbar_ind in mpl.toolbar_items){\n",
" var name = mpl.toolbar_items[toolbar_ind][0];\n",
" var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
" var image = mpl.toolbar_items[toolbar_ind][2];\n",
" var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
"\n",
" if (!name) { continue; };\n",
"\n",
" var button = $('');\n",
" button.click(method_name, toolbar_event);\n",
" button.mouseover(tooltip, toolbar_mouse_event);\n",
" nav_element.append(button);\n",
" }\n",
"\n",
" // Add the status bar.\n",
" var status_bar = $('');\n",
" nav_element.append(status_bar);\n",
" this.message = status_bar[0];\n",
"\n",
" // Add the close button to the window.\n",
" var buttongrp = $('');\n",
" var button = $('');\n",
" button.click(function (evt) { fig.handle_close(fig, {}); } );\n",
" button.mouseover('Stop Interaction', toolbar_mouse_event);\n",
" buttongrp.append(button);\n",
" var titlebar = this.root.find($('.ui-dialog-titlebar'));\n",
" titlebar.prepend(buttongrp);\n",
"}\n",
"\n",
"mpl.figure.prototype._root_extra_style = function(el){\n",
" var fig = this\n",
" el.on(\"remove\", function(){\n",
"\tfig.close_ws(fig, {});\n",
" });\n",
"}\n",
"\n",
"mpl.figure.prototype._canvas_extra_style = function(el){\n",
" // this is important to make the div 'focusable\n",
" el.attr('tabindex', 0)\n",
" // reach out to IPython and tell the keyboard manager to turn it's self\n",
" // off when our div gets focus\n",
"\n",
" // location in version 3\n",
" if (IPython.notebook.keyboard_manager) {\n",
" IPython.notebook.keyboard_manager.register_events(el);\n",
" }\n",
" else {\n",
" // location in version 2\n",
" IPython.keyboard_manager.register_events(el);\n",
" }\n",
"\n",
"}\n",
"\n",
"mpl.figure.prototype._key_event_extra = function(event, name) {\n",
" var manager = IPython.notebook.keyboard_manager;\n",
" if (!manager)\n",
" manager = IPython.keyboard_manager;\n",
"\n",
" // Check for shift+enter\n",
" if (event.shiftKey && event.which == 13) {\n",
" this.canvas_div.blur();\n",
" event.shiftKey = false;\n",
" // Send a \"J\" for go to next cell\n",
" event.which = 74;\n",
" event.keyCode = 74;\n",
" manager.command_mode();\n",
" manager.handle_keydown(event);\n",
" }\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_save = function(fig, msg) {\n",
" fig.ondownload(fig, null);\n",
"}\n",
"\n",
"\n",
"mpl.find_output_cell = function(html_output) {\n",
" // Return the cell and output element which can be found *uniquely* in the notebook.\n",
" // Note - this is a bit hacky, but it is done because the \"notebook_saving.Notebook\"\n",
" // IPython event is triggered only after the cells have been serialised, which for\n",
" // our purposes (turning an active figure into a static one), is too late.\n",
" var cells = IPython.notebook.get_cells();\n",
" var ncells = cells.length;\n",
" for (var i=0; i= 3 moved mimebundle to data attribute of output\n",
" data = data.data;\n",
" }\n",
" if (data['text/html'] == html_output) {\n",
" return [cell, data, j];\n",
" }\n",
" }\n",
" }\n",
" }\n",
"}\n",
"\n",
"// Register the function which deals with the matplotlib target/channel.\n",
"// The kernel may be null if the page has been refreshed.\n",
"if (IPython.notebook.kernel != null) {\n",
" IPython.notebook.kernel.comm_manager.register_target('matplotlib', mpl.mpl_figure_comm);\n",
"}\n"
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/html": [
""
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"colors = 'RGB'\n",
"c = colors.find('R')\n",
"T = bayes_thresh(im_noise)\n",
"\n",
"plt.figure(figsize=(10,10))\n",
"gs = gridspec.GridSpec(3,3)\n",
"gs.update(left=0.1, right=0.95, wspace=0.05, hspace=.2, top=0.95)\n",
"for j in range(J):\n",
" for b, band in enumerate(['LH', 'HL', 'HH']): \n",
" if b == 0: \n",
" ax = plt.subplot(gs[j,b])\n",
" ax.yaxis.set_major_formatter(FormatStrFormatter('%.2f'))\n",
" ax0 = ax \n",
" else:\n",
" ax = plt.subplot(gs[j,b])\n",
" plt.setp(ax.get_yticklabels(), visible=False)\n",
" \n",
" ax.plot(Ts[j,b,c], mses[j,b,c])\n",
" if j == 0 and b == 1:\n",
" ax.axhline(y=T_mse[j,b,c], ls='--',c='r', \n",
" label='MSE for bayes calculated thresh')\n",
" ax.legend()\n",
" else:\n",
" ax.axhline(y=T_mse[j,b,c], ls='--',c='r')\n",
" ax.axvline(x=T[j,b,c], ls='--',c='r')\n",
" ax.set_title('Scale {} {}'.format(J-j, band))\n",
" ax.set_xlim(0, max(Ts[j,b,c,-1], T[j,b,c]*1.1))\n",
" "
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Very interesting, the bayes calculated threshold is very near the true minimum for almost all the subbands!"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Convexity Test\n",
"Test the above assertion, that minimizing each threshold independently results in minimizing the overall MSE. First, search over the above plots for\n",
"$$T^{*} = \\text{argmin}_{T} MSE(x, \\hat{x})$$"
]
},
{
"cell_type": "code",
"execution_count": 14,
"metadata": {
"ExecuteTime": {
"end_time": "2018-12-03T16:52:38.326838Z",
"start_time": "2018-12-03T16:52:37.821060Z"
},
"scrolled": true
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Bayes Shrink MSE is 85.4250447125913\n",
"Min shrink MSE is 82.9359235938321\n"
]
}
],
"source": [
"# Calculate T*\n",
"mse_min = np.argmin(mses, axis=-1)\n",
"T_star = np.zeros_like(T)\n",
"for j in range(J):\n",
" for b in range(3):\n",
" for c in range(C):\n",
" T_star[j,b,c] = Ts[j,b,c, mse_min[j,b,c]]\n",
" \n",
"# Calculate the MSE with T* vs with the bayes estimated thresholds\n",
"coeffs = pywt.wavedec2(im_noise, wave, axes=(0,1), level=J)\n",
"x_hat_min = pywt.waverec2(shrink_coeffs(coeffs, T_star), wave, axes=(0,1))\n",
"x_hat_bayes = pywt.waverec2(shrink_coeffs(coeffs, T), wave, axes=(0,1))\n",
"print('Bayes Shrink MSE is {}'.format(np.sum((im-x_hat_bayes)**2/im.size)))\n",
"print('Min shrink MSE is {}'.format(np.sum((im-x_hat_min)**2/im.size))) "
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"We haven't really proven it's the global minimum, but we have proven that it is at least better than using the bayes shrink thresholds."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# Find threshold by gradients"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Finite Differences\n",
"The above curve for $L(T)$ looks quite smooth, so it seems plausible that the gradients are smooth. Let us test these via finite differences"
]
},
{
"cell_type": "code",
"execution_count": 15,
"metadata": {
"ExecuteTime": {
"end_time": "2018-12-03T16:52:38.386311Z",
"start_time": "2018-12-03T16:52:38.328507Z"
}
},
"outputs": [
{
"data": {
"application/javascript": [
"/* Put everything inside the global mpl namespace */\n",
"window.mpl = {};\n",
"\n",
"\n",
"mpl.get_websocket_type = function() {\n",
" if (typeof(WebSocket) !== 'undefined') {\n",
" return WebSocket;\n",
" } else if (typeof(MozWebSocket) !== 'undefined') {\n",
" return MozWebSocket;\n",
" } else {\n",
" alert('Your browser does not have WebSocket support.' +\n",
" 'Please try Chrome, Safari or Firefox ≥ 6. ' +\n",
" 'Firefox 4 and 5 are also supported but you ' +\n",
" 'have to enable WebSockets in about:config.');\n",
" };\n",
"}\n",
"\n",
"mpl.figure = function(figure_id, websocket, ondownload, parent_element) {\n",
" this.id = figure_id;\n",
"\n",
" this.ws = websocket;\n",
"\n",
" this.supports_binary = (this.ws.binaryType != undefined);\n",
"\n",
" if (!this.supports_binary) {\n",
" var warnings = document.getElementById(\"mpl-warnings\");\n",
" if (warnings) {\n",
" warnings.style.display = 'block';\n",
" warnings.textContent = (\n",
" \"This browser does not support binary websocket messages. \" +\n",
" \"Performance may be slow.\");\n",
" }\n",
" }\n",
"\n",
" this.imageObj = new Image();\n",
"\n",
" this.context = undefined;\n",
" this.message = undefined;\n",
" this.canvas = undefined;\n",
" this.rubberband_canvas = undefined;\n",
" this.rubberband_context = undefined;\n",
" this.format_dropdown = undefined;\n",
"\n",
" this.image_mode = 'full';\n",
"\n",
" this.root = $('');\n",
" this._root_extra_style(this.root)\n",
" this.root.attr('style', 'display: inline-block');\n",
"\n",
" $(parent_element).append(this.root);\n",
"\n",
" this._init_header(this);\n",
" this._init_canvas(this);\n",
" this._init_toolbar(this);\n",
"\n",
" var fig = this;\n",
"\n",
" this.waiting = false;\n",
"\n",
" this.ws.onopen = function () {\n",
" fig.send_message(\"supports_binary\", {value: fig.supports_binary});\n",
" fig.send_message(\"send_image_mode\", {});\n",
" if (mpl.ratio != 1) {\n",
" fig.send_message(\"set_dpi_ratio\", {'dpi_ratio': mpl.ratio});\n",
" }\n",
" fig.send_message(\"refresh\", {});\n",
" }\n",
"\n",
" this.imageObj.onload = function() {\n",
" if (fig.image_mode == 'full') {\n",
" // Full images could contain transparency (where diff images\n",
" // almost always do), so we need to clear the canvas so that\n",
" // there is no ghosting.\n",
" fig.context.clearRect(0, 0, fig.canvas.width, fig.canvas.height);\n",
" }\n",
" fig.context.drawImage(fig.imageObj, 0, 0);\n",
" };\n",
"\n",
" this.imageObj.onunload = function() {\n",
" fig.ws.close();\n",
" }\n",
"\n",
" this.ws.onmessage = this._make_on_message_function(this);\n",
"\n",
" this.ondownload = ondownload;\n",
"}\n",
"\n",
"mpl.figure.prototype._init_header = function() {\n",
" var titlebar = $(\n",
" '');\n",
" var titletext = $(\n",
" '');\n",
" titlebar.append(titletext)\n",
" this.root.append(titlebar);\n",
" this.header = titletext[0];\n",
"}\n",
"\n",
"\n",
"\n",
"mpl.figure.prototype._canvas_extra_style = function(canvas_div) {\n",
"\n",
"}\n",
"\n",
"\n",
"mpl.figure.prototype._root_extra_style = function(canvas_div) {\n",
"\n",
"}\n",
"\n",
"mpl.figure.prototype._init_canvas = function() {\n",
" var fig = this;\n",
"\n",
" var canvas_div = $('');\n",
"\n",
" canvas_div.attr('style', 'position: relative; clear: both; outline: 0');\n",
"\n",
" function canvas_keyboard_event(event) {\n",
" return fig.key_event(event, event['data']);\n",
" }\n",
"\n",
" canvas_div.keydown('key_press', canvas_keyboard_event);\n",
" canvas_div.keyup('key_release', canvas_keyboard_event);\n",
" this.canvas_div = canvas_div\n",
" this._canvas_extra_style(canvas_div)\n",
" this.root.append(canvas_div);\n",
"\n",
" var canvas = $('');\n",
" canvas.addClass('mpl-canvas');\n",
" canvas.attr('style', \"left: 0; top: 0; z-index: 0; outline: 0\")\n",
"\n",
" this.canvas = canvas[0];\n",
" this.context = canvas[0].getContext(\"2d\");\n",
"\n",
" var backingStore = this.context.backingStorePixelRatio ||\n",
"\tthis.context.webkitBackingStorePixelRatio ||\n",
"\tthis.context.mozBackingStorePixelRatio ||\n",
"\tthis.context.msBackingStorePixelRatio ||\n",
"\tthis.context.oBackingStorePixelRatio ||\n",
"\tthis.context.backingStorePixelRatio || 1;\n",
"\n",
" mpl.ratio = (window.devicePixelRatio || 1) / backingStore;\n",
"\n",
" var rubberband = $('');\n",
" rubberband.attr('style', \"position: absolute; left: 0; top: 0; z-index: 1;\")\n",
"\n",
" var pass_mouse_events = true;\n",
"\n",
" canvas_div.resizable({\n",
" start: function(event, ui) {\n",
" pass_mouse_events = false;\n",
" },\n",
" resize: function(event, ui) {\n",
" fig.request_resize(ui.size.width, ui.size.height);\n",
" },\n",
" stop: function(event, ui) {\n",
" pass_mouse_events = true;\n",
" fig.request_resize(ui.size.width, ui.size.height);\n",
" },\n",
" });\n",
"\n",
" function mouse_event_fn(event) {\n",
" if (pass_mouse_events)\n",
" return fig.mouse_event(event, event['data']);\n",
" }\n",
"\n",
" rubberband.mousedown('button_press', mouse_event_fn);\n",
" rubberband.mouseup('button_release', mouse_event_fn);\n",
" // Throttle sequential mouse events to 1 every 20ms.\n",
" rubberband.mousemove('motion_notify', mouse_event_fn);\n",
"\n",
" rubberband.mouseenter('figure_enter', mouse_event_fn);\n",
" rubberband.mouseleave('figure_leave', mouse_event_fn);\n",
"\n",
" canvas_div.on(\"wheel\", function (event) {\n",
" event = event.originalEvent;\n",
" event['data'] = 'scroll'\n",
" if (event.deltaY < 0) {\n",
" event.step = 1;\n",
" } else {\n",
" event.step = -1;\n",
" }\n",
" mouse_event_fn(event);\n",
" });\n",
"\n",
" canvas_div.append(canvas);\n",
" canvas_div.append(rubberband);\n",
"\n",
" this.rubberband = rubberband;\n",
" this.rubberband_canvas = rubberband[0];\n",
" this.rubberband_context = rubberband[0].getContext(\"2d\");\n",
" this.rubberband_context.strokeStyle = \"#000000\";\n",
"\n",
" this._resize_canvas = function(width, height) {\n",
" // Keep the size of the canvas, canvas container, and rubber band\n",
" // canvas in synch.\n",
" canvas_div.css('width', width)\n",
" canvas_div.css('height', height)\n",
"\n",
" canvas.attr('width', width * mpl.ratio);\n",
" canvas.attr('height', height * mpl.ratio);\n",
" canvas.attr('style', 'width: ' + width + 'px; height: ' + height + 'px;');\n",
"\n",
" rubberband.attr('width', width);\n",
" rubberband.attr('height', height);\n",
" }\n",
"\n",
" // Set the figure to an initial 600x600px, this will subsequently be updated\n",
" // upon first draw.\n",
" this._resize_canvas(600, 600);\n",
"\n",
" // Disable right mouse context menu.\n",
" $(this.rubberband_canvas).bind(\"contextmenu\",function(e){\n",
" return false;\n",
" });\n",
"\n",
" function set_focus () {\n",
" canvas.focus();\n",
" canvas_div.focus();\n",
" }\n",
"\n",
" window.setTimeout(set_focus, 100);\n",
"}\n",
"\n",
"mpl.figure.prototype._init_toolbar = function() {\n",
" var fig = this;\n",
"\n",
" var nav_element = $('')\n",
" nav_element.attr('style', 'width: 100%');\n",
" this.root.append(nav_element);\n",
"\n",
" // Define a callback function for later on.\n",
" function toolbar_event(event) {\n",
" return fig.toolbar_button_onclick(event['data']);\n",
" }\n",
" function toolbar_mouse_event(event) {\n",
" return fig.toolbar_button_onmouseover(event['data']);\n",
" }\n",
"\n",
" for(var toolbar_ind in mpl.toolbar_items) {\n",
" var name = mpl.toolbar_items[toolbar_ind][0];\n",
" var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
" var image = mpl.toolbar_items[toolbar_ind][2];\n",
" var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
"\n",
" if (!name) {\n",
" // put a spacer in here.\n",
" continue;\n",
" }\n",
" var button = $('');\n",
" button.addClass('ui-button ui-widget ui-state-default ui-corner-all ' +\n",
" 'ui-button-icon-only');\n",
" button.attr('role', 'button');\n",
" button.attr('aria-disabled', 'false');\n",
" button.click(method_name, toolbar_event);\n",
" button.mouseover(tooltip, toolbar_mouse_event);\n",
"\n",
" var icon_img = $('');\n",
" icon_img.addClass('ui-button-icon-primary ui-icon');\n",
" icon_img.addClass(image);\n",
" icon_img.addClass('ui-corner-all');\n",
"\n",
" var tooltip_span = $('');\n",
" tooltip_span.addClass('ui-button-text');\n",
" tooltip_span.html(tooltip);\n",
"\n",
" button.append(icon_img);\n",
" button.append(tooltip_span);\n",
"\n",
" nav_element.append(button);\n",
" }\n",
"\n",
" var fmt_picker_span = $('');\n",
"\n",
" var fmt_picker = $('');\n",
" fmt_picker.addClass('mpl-toolbar-option ui-widget ui-widget-content');\n",
" fmt_picker_span.append(fmt_picker);\n",
" nav_element.append(fmt_picker_span);\n",
" this.format_dropdown = fmt_picker[0];\n",
"\n",
" for (var ind in mpl.extensions) {\n",
" var fmt = mpl.extensions[ind];\n",
" var option = $(\n",
" '', {selected: fmt === mpl.default_extension}).html(fmt);\n",
" fmt_picker.append(option)\n",
" }\n",
"\n",
" // Add hover states to the ui-buttons\n",
" $( \".ui-button\" ).hover(\n",
" function() { $(this).addClass(\"ui-state-hover\");},\n",
" function() { $(this).removeClass(\"ui-state-hover\");}\n",
" );\n",
"\n",
" var status_bar = $('');\n",
" nav_element.append(status_bar);\n",
" this.message = status_bar[0];\n",
"}\n",
"\n",
"mpl.figure.prototype.request_resize = function(x_pixels, y_pixels) {\n",
" // Request matplotlib to resize the figure. Matplotlib will then trigger a resize in the client,\n",
" // which will in turn request a refresh of the image.\n",
" this.send_message('resize', {'width': x_pixels, 'height': y_pixels});\n",
"}\n",
"\n",
"mpl.figure.prototype.send_message = function(type, properties) {\n",
" properties['type'] = type;\n",
" properties['figure_id'] = this.id;\n",
" this.ws.send(JSON.stringify(properties));\n",
"}\n",
"\n",
"mpl.figure.prototype.send_draw_message = function() {\n",
" if (!this.waiting) {\n",
" this.waiting = true;\n",
" this.ws.send(JSON.stringify({type: \"draw\", figure_id: this.id}));\n",
" }\n",
"}\n",
"\n",
"\n",
"mpl.figure.prototype.handle_save = function(fig, msg) {\n",
" var format_dropdown = fig.format_dropdown;\n",
" var format = format_dropdown.options[format_dropdown.selectedIndex].value;\n",
" fig.ondownload(fig, format);\n",
"}\n",
"\n",
"\n",
"mpl.figure.prototype.handle_resize = function(fig, msg) {\n",
" var size = msg['size'];\n",
" if (size[0] != fig.canvas.width || size[1] != fig.canvas.height) {\n",
" fig._resize_canvas(size[0], size[1]);\n",
" fig.send_message(\"refresh\", {});\n",
" };\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_rubberband = function(fig, msg) {\n",
" var x0 = msg['x0'] / mpl.ratio;\n",
" var y0 = (fig.canvas.height - msg['y0']) / mpl.ratio;\n",
" var x1 = msg['x1'] / mpl.ratio;\n",
" var y1 = (fig.canvas.height - msg['y1']) / mpl.ratio;\n",
" x0 = Math.floor(x0) + 0.5;\n",
" y0 = Math.floor(y0) + 0.5;\n",
" x1 = Math.floor(x1) + 0.5;\n",
" y1 = Math.floor(y1) + 0.5;\n",
" var min_x = Math.min(x0, x1);\n",
" var min_y = Math.min(y0, y1);\n",
" var width = Math.abs(x1 - x0);\n",
" var height = Math.abs(y1 - y0);\n",
"\n",
" fig.rubberband_context.clearRect(\n",
" 0, 0, fig.canvas.width, fig.canvas.height);\n",
"\n",
" fig.rubberband_context.strokeRect(min_x, min_y, width, height);\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_figure_label = function(fig, msg) {\n",
" // Updates the figure title.\n",
" fig.header.textContent = msg['label'];\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_cursor = function(fig, msg) {\n",
" var cursor = msg['cursor'];\n",
" switch(cursor)\n",
" {\n",
" case 0:\n",
" cursor = 'pointer';\n",
" break;\n",
" case 1:\n",
" cursor = 'default';\n",
" break;\n",
" case 2:\n",
" cursor = 'crosshair';\n",
" break;\n",
" case 3:\n",
" cursor = 'move';\n",
" break;\n",
" }\n",
" fig.rubberband_canvas.style.cursor = cursor;\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_message = function(fig, msg) {\n",
" fig.message.textContent = msg['message'];\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_draw = function(fig, msg) {\n",
" // Request the server to send over a new figure.\n",
" fig.send_draw_message();\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_image_mode = function(fig, msg) {\n",
" fig.image_mode = msg['mode'];\n",
"}\n",
"\n",
"mpl.figure.prototype.updated_canvas_event = function() {\n",
" // Called whenever the canvas gets updated.\n",
" this.send_message(\"ack\", {});\n",
"}\n",
"\n",
"// A function to construct a web socket function for onmessage handling.\n",
"// Called in the figure constructor.\n",
"mpl.figure.prototype._make_on_message_function = function(fig) {\n",
" return function socket_on_message(evt) {\n",
" if (evt.data instanceof Blob) {\n",
" /* FIXME: We get \"Resource interpreted as Image but\n",
" * transferred with MIME type text/plain:\" errors on\n",
" * Chrome. But how to set the MIME type? It doesn't seem\n",
" * to be part of the websocket stream */\n",
" evt.data.type = \"image/png\";\n",
"\n",
" /* Free the memory for the previous frames */\n",
" if (fig.imageObj.src) {\n",
" (window.URL || window.webkitURL).revokeObjectURL(\n",
" fig.imageObj.src);\n",
" }\n",
"\n",
" fig.imageObj.src = (window.URL || window.webkitURL).createObjectURL(\n",
" evt.data);\n",
" fig.updated_canvas_event();\n",
" fig.waiting = false;\n",
" return;\n",
" }\n",
" else if (typeof evt.data === 'string' && evt.data.slice(0, 21) == \"data:image/png;base64\") {\n",
" fig.imageObj.src = evt.data;\n",
" fig.updated_canvas_event();\n",
" fig.waiting = false;\n",
" return;\n",
" }\n",
"\n",
" var msg = JSON.parse(evt.data);\n",
" var msg_type = msg['type'];\n",
"\n",
" // Call the \"handle_{type}\" callback, which takes\n",
" // the figure and JSON message as its only arguments.\n",
" try {\n",
" var callback = fig[\"handle_\" + msg_type];\n",
" } catch (e) {\n",
" console.log(\"No handler for the '\" + msg_type + \"' message type: \", msg);\n",
" return;\n",
" }\n",
"\n",
" if (callback) {\n",
" try {\n",
" // console.log(\"Handling '\" + msg_type + \"' message: \", msg);\n",
" callback(fig, msg);\n",
" } catch (e) {\n",
" console.log(\"Exception inside the 'handler_\" + msg_type + \"' callback:\", e, e.stack, msg);\n",
" }\n",
" }\n",
" };\n",
"}\n",
"\n",
"// from http://stackoverflow.com/questions/1114465/getting-mouse-location-in-canvas\n",
"mpl.findpos = function(e) {\n",
" //this section is from http://www.quirksmode.org/js/events_properties.html\n",
" var targ;\n",
" if (!e)\n",
" e = window.event;\n",
" if (e.target)\n",
" targ = e.target;\n",
" else if (e.srcElement)\n",
" targ = e.srcElement;\n",
" if (targ.nodeType == 3) // defeat Safari bug\n",
" targ = targ.parentNode;\n",
"\n",
" // jQuery normalizes the pageX and pageY\n",
" // pageX,Y are the mouse positions relative to the document\n",
" // offset() returns the position of the element relative to the document\n",
" var x = e.pageX - $(targ).offset().left;\n",
" var y = e.pageY - $(targ).offset().top;\n",
"\n",
" return {\"x\": x, \"y\": y};\n",
"};\n",
"\n",
"/*\n",
" * return a copy of an object with only non-object keys\n",
" * we need this to avoid circular references\n",
" * http://stackoverflow.com/a/24161582/3208463\n",
" */\n",
"function simpleKeys (original) {\n",
" return Object.keys(original).reduce(function (obj, key) {\n",
" if (typeof original[key] !== 'object')\n",
" obj[key] = original[key]\n",
" return obj;\n",
" }, {});\n",
"}\n",
"\n",
"mpl.figure.prototype.mouse_event = function(event, name) {\n",
" var canvas_pos = mpl.findpos(event)\n",
"\n",
" if (name === 'button_press')\n",
" {\n",
" this.canvas.focus();\n",
" this.canvas_div.focus();\n",
" }\n",
"\n",
" var x = canvas_pos.x * mpl.ratio;\n",
" var y = canvas_pos.y * mpl.ratio;\n",
"\n",
" this.send_message(name, {x: x, y: y, button: event.button,\n",
" step: event.step,\n",
" guiEvent: simpleKeys(event)});\n",
"\n",
" /* This prevents the web browser from automatically changing to\n",
" * the text insertion cursor when the button is pressed. We want\n",
" * to control all of the cursor setting manually through the\n",
" * 'cursor' event from matplotlib */\n",
" event.preventDefault();\n",
" return false;\n",
"}\n",
"\n",
"mpl.figure.prototype._key_event_extra = function(event, name) {\n",
" // Handle any extra behaviour associated with a key event\n",
"}\n",
"\n",
"mpl.figure.prototype.key_event = function(event, name) {\n",
"\n",
" // Prevent repeat events\n",
" if (name == 'key_press')\n",
" {\n",
" if (event.which === this._key)\n",
" return;\n",
" else\n",
" this._key = event.which;\n",
" }\n",
" if (name == 'key_release')\n",
" this._key = null;\n",
"\n",
" var value = '';\n",
" if (event.ctrlKey && event.which != 17)\n",
" value += \"ctrl+\";\n",
" if (event.altKey && event.which != 18)\n",
" value += \"alt+\";\n",
" if (event.shiftKey && event.which != 16)\n",
" value += \"shift+\";\n",
"\n",
" value += 'k';\n",
" value += event.which.toString();\n",
"\n",
" this._key_event_extra(event, name);\n",
"\n",
" this.send_message(name, {key: value,\n",
" guiEvent: simpleKeys(event)});\n",
" return false;\n",
"}\n",
"\n",
"mpl.figure.prototype.toolbar_button_onclick = function(name) {\n",
" if (name == 'download') {\n",
" this.handle_save(this, null);\n",
" } else {\n",
" this.send_message(\"toolbar_button\", {name: name});\n",
" }\n",
"};\n",
"\n",
"mpl.figure.prototype.toolbar_button_onmouseover = function(tooltip) {\n",
" this.message.textContent = tooltip;\n",
"};\n",
"mpl.toolbar_items = [[\"Home\", \"Reset original view\", \"fa fa-home icon-home\", \"home\"], [\"Back\", \"Back to previous view\", \"fa fa-arrow-left icon-arrow-left\", \"back\"], [\"Forward\", \"Forward to next view\", \"fa fa-arrow-right icon-arrow-right\", \"forward\"], [\"\", \"\", \"\", \"\"], [\"Pan\", \"Pan axes with left mouse, zoom with right\", \"fa fa-arrows icon-move\", \"pan\"], [\"Zoom\", \"Zoom to rectangle\", \"fa fa-square-o icon-check-empty\", \"zoom\"], [\"\", \"\", \"\", \"\"], [\"Download\", \"Download plot\", \"fa fa-floppy-o icon-save\", \"download\"]];\n",
"\n",
"mpl.extensions = [\"eps\", \"jpeg\", \"pdf\", \"png\", \"ps\", \"raw\", \"svg\", \"tif\"];\n",
"\n",
"mpl.default_extension = \"png\";var comm_websocket_adapter = function(comm) {\n",
" // Create a \"websocket\"-like object which calls the given IPython comm\n",
" // object with the appropriate methods. Currently this is a non binary\n",
" // socket, so there is still some room for performance tuning.\n",
" var ws = {};\n",
"\n",
" ws.close = function() {\n",
" comm.close()\n",
" };\n",
" ws.send = function(m) {\n",
" //console.log('sending', m);\n",
" comm.send(m);\n",
" };\n",
" // Register the callback with on_msg.\n",
" comm.on_msg(function(msg) {\n",
" //console.log('receiving', msg['content']['data'], msg);\n",
" // Pass the mpl event to the overridden (by mpl) onmessage function.\n",
" ws.onmessage(msg['content']['data'])\n",
" });\n",
" return ws;\n",
"}\n",
"\n",
"mpl.mpl_figure_comm = function(comm, msg) {\n",
" // This is the function which gets called when the mpl process\n",
" // starts-up an IPython Comm through the \"matplotlib\" channel.\n",
"\n",
" var id = msg.content.data.id;\n",
" // Get hold of the div created by the display call when the Comm\n",
" // socket was opened in Python.\n",
" var element = $(\"#\" + id);\n",
" var ws_proxy = comm_websocket_adapter(comm)\n",
"\n",
" function ondownload(figure, format) {\n",
" window.open(figure.imageObj.src);\n",
" }\n",
"\n",
" var fig = new mpl.figure(id, ws_proxy,\n",
" ondownload,\n",
" element.get(0));\n",
"\n",
" // Call onopen now - mpl needs it, as it is assuming we've passed it a real\n",
" // web socket which is closed, not our websocket->open comm proxy.\n",
" ws_proxy.onopen();\n",
"\n",
" fig.parent_element = element.get(0);\n",
" fig.cell_info = mpl.find_output_cell(\"\");\n",
" if (!fig.cell_info) {\n",
" console.error(\"Failed to find cell for figure\", id, fig);\n",
" return;\n",
" }\n",
"\n",
" var output_index = fig.cell_info[2]\n",
" var cell = fig.cell_info[0];\n",
"\n",
"};\n",
"\n",
"mpl.figure.prototype.handle_close = function(fig, msg) {\n",
" var width = fig.canvas.width/mpl.ratio\n",
" fig.root.unbind('remove')\n",
"\n",
" // Update the output cell to use the data from the current canvas.\n",
" fig.push_to_output();\n",
" var dataURL = fig.canvas.toDataURL();\n",
" // Re-enable the keyboard manager in IPython - without this line, in FF,\n",
" // the notebook keyboard shortcuts fail.\n",
" IPython.keyboard_manager.enable()\n",
" $(fig.parent_element).html('');\n",
" fig.close_ws(fig, msg);\n",
"}\n",
"\n",
"mpl.figure.prototype.close_ws = function(fig, msg){\n",
" fig.send_message('closing', msg);\n",
" // fig.ws.close()\n",
"}\n",
"\n",
"mpl.figure.prototype.push_to_output = function(remove_interactive) {\n",
" // Turn the data on the canvas into data in the output cell.\n",
" var width = this.canvas.width/mpl.ratio\n",
" var dataURL = this.canvas.toDataURL();\n",
" this.cell_info[1]['text/html'] = '';\n",
"}\n",
"\n",
"mpl.figure.prototype.updated_canvas_event = function() {\n",
" // Tell IPython that the notebook contents must change.\n",
" IPython.notebook.set_dirty(true);\n",
" this.send_message(\"ack\", {});\n",
" var fig = this;\n",
" // Wait a second, then push the new image to the DOM so\n",
" // that it is saved nicely (might be nice to debounce this).\n",
" setTimeout(function () { fig.push_to_output() }, 1000);\n",
"}\n",
"\n",
"mpl.figure.prototype._init_toolbar = function() {\n",
" var fig = this;\n",
"\n",
" var nav_element = $('')\n",
" nav_element.attr('style', 'width: 100%');\n",
" this.root.append(nav_element);\n",
"\n",
" // Define a callback function for later on.\n",
" function toolbar_event(event) {\n",
" return fig.toolbar_button_onclick(event['data']);\n",
" }\n",
" function toolbar_mouse_event(event) {\n",
" return fig.toolbar_button_onmouseover(event['data']);\n",
" }\n",
"\n",
" for(var toolbar_ind in mpl.toolbar_items){\n",
" var name = mpl.toolbar_items[toolbar_ind][0];\n",
" var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
" var image = mpl.toolbar_items[toolbar_ind][2];\n",
" var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
"\n",
" if (!name) { continue; };\n",
"\n",
" var button = $('');\n",
" button.click(method_name, toolbar_event);\n",
" button.mouseover(tooltip, toolbar_mouse_event);\n",
" nav_element.append(button);\n",
" }\n",
"\n",
" // Add the status bar.\n",
" var status_bar = $('');\n",
" nav_element.append(status_bar);\n",
" this.message = status_bar[0];\n",
"\n",
" // Add the close button to the window.\n",
" var buttongrp = $('');\n",
" var button = $('');\n",
" button.click(function (evt) { fig.handle_close(fig, {}); } );\n",
" button.mouseover('Stop Interaction', toolbar_mouse_event);\n",
" buttongrp.append(button);\n",
" var titlebar = this.root.find($('.ui-dialog-titlebar'));\n",
" titlebar.prepend(buttongrp);\n",
"}\n",
"\n",
"mpl.figure.prototype._root_extra_style = function(el){\n",
" var fig = this\n",
" el.on(\"remove\", function(){\n",
"\tfig.close_ws(fig, {});\n",
" });\n",
"}\n",
"\n",
"mpl.figure.prototype._canvas_extra_style = function(el){\n",
" // this is important to make the div 'focusable\n",
" el.attr('tabindex', 0)\n",
" // reach out to IPython and tell the keyboard manager to turn it's self\n",
" // off when our div gets focus\n",
"\n",
" // location in version 3\n",
" if (IPython.notebook.keyboard_manager) {\n",
" IPython.notebook.keyboard_manager.register_events(el);\n",
" }\n",
" else {\n",
" // location in version 2\n",
" IPython.keyboard_manager.register_events(el);\n",
" }\n",
"\n",
"}\n",
"\n",
"mpl.figure.prototype._key_event_extra = function(event, name) {\n",
" var manager = IPython.notebook.keyboard_manager;\n",
" if (!manager)\n",
" manager = IPython.keyboard_manager;\n",
"\n",
" // Check for shift+enter\n",
" if (event.shiftKey && event.which == 13) {\n",
" this.canvas_div.blur();\n",
" event.shiftKey = false;\n",
" // Send a \"J\" for go to next cell\n",
" event.which = 74;\n",
" event.keyCode = 74;\n",
" manager.command_mode();\n",
" manager.handle_keydown(event);\n",
" }\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_save = function(fig, msg) {\n",
" fig.ondownload(fig, null);\n",
"}\n",
"\n",
"\n",
"mpl.find_output_cell = function(html_output) {\n",
" // Return the cell and output element which can be found *uniquely* in the notebook.\n",
" // Note - this is a bit hacky, but it is done because the \"notebook_saving.Notebook\"\n",
" // IPython event is triggered only after the cells have been serialised, which for\n",
" // our purposes (turning an active figure into a static one), is too late.\n",
" var cells = IPython.notebook.get_cells();\n",
" var ncells = cells.length;\n",
" for (var i=0; i= 3 moved mimebundle to data attribute of output\n",
" data = data.data;\n",
" }\n",
" if (data['text/html'] == html_output) {\n",
" return [cell, data, j];\n",
" }\n",
" }\n",
" }\n",
" }\n",
"}\n",
"\n",
"// Register the function which deals with the matplotlib target/channel.\n",
"// The kernel may be null if the page has been refreshed.\n",
"if (IPython.notebook.kernel != null) {\n",
" IPython.notebook.kernel.comm_manager.register_target('matplotlib', mpl.mpl_figure_comm);\n",
"}\n"
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/html": [
""
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"fig, ax = plt.subplots(1)\n",
"ax.yaxis.set_major_formatter(FormatStrFormatter('%.2f'))\n",
"ax.plot(Ts[0,0,1], mses[0,0,1])\n",
"ax.set_title('LH subband for scale 3')\n",
"ax.set_xlabel('T')\n",
"ax.set_ylabel('L(T)')\n",
"plt.tight_layout()"
]
},
{
"cell_type": "code",
"execution_count": 16,
"metadata": {
"ExecuteTime": {
"end_time": "2018-12-03T16:52:40.393872Z",
"start_time": "2018-12-03T16:52:38.388034Z"
}
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Gradient at T=5 is -0.0592\n",
"Gradient at T=20 is 0.0342\n"
]
}
],
"source": [
"# Calculate the mean removed image\n",
"coeffs = pywt.wavedec2(im_noise, wave, axes=(0,1), level=J)\n",
"coeffs_zero = pywt.wavedec2(np.zeros_like(im_noise), wave, axes=(0,1), \n",
" level=J)\n",
"coeffs_zero[0] = np.copy(coeffs[0])\n",
"μ = pywt.waverec2(coeffs_zero, wave, axes=(0,1))\n",
"im2 = im - μ\n",
"\n",
"def F(t):\n",
" coeffs = pywt.wavedec2(im_noise, wave, axes=(0,1), level=J)\n",
" coeffs_zero = pywt.wavedec2(np.zeros_like(im_noise), wave, axes=(0,1), \n",
" level=J)\n",
" coeffs_zero[1][0][:,:,1] = shrink(coeffs[1][0][:,:,1], t)\n",
" x_hat = pywt.waverec2(coeffs_zero, wave, axes=(0,1))\n",
" return x_hat\n",
"\n",
"def L(t): \n",
" return np.mean((im2 - F(t))**2)\n",
" \n",
"def dLdt(T, dt=.1): \n",
" return (L(T+dt) - L(T-dt))/(2*dt)\n",
"\n",
"print('Gradient at T=5 is {:.4f}'.format(dLdt(5)))\n",
"print('Gradient at T=20 is {:.4f}'.format(dLdt(20)))"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"This matches up with the plotted curve"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Manual backpropagation"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Analytically what are the gradients?\n",
"$$\n",
"L=\\frac{1}{MN}(\\hat{X}-X)^2\n",
"$$\n",
"\n",
"by modifying $T$? \n",
"\n",
"- It is trivial to calculate $\\frac{dL}{d\\hat{X}}$. \n",
"- Calculating $\\frac{d\\hat{X}}{dw}$ is not too difficult - simply need to take the forward wavelet transform with time reversed synthesis filters as the analysis filters. As we used orthogonal wavelets, this is already the case.\n",
"- Calculating $\\frac{dw}{dT}$ is also trivial - I have done this already. To remind us, if:\n",
"\n",
"$$\n",
"w = \\eta_T(x) = x \\cdot \\frac{max(|x|-t, 0)}{|x|}\n",
"$$\n",
"\n",
"then \n",
"$$\n",
"\\frac{dw}{dT} = -\\text{sign} (x)\\mathbb{1}(|x| - t > 0) = -\\text{sign}(w) \\in \\{-1, 0, 1 \\}\n",
"$$\n",
"\n",
"and \n",
"$$\n",
"\\frac{dw}{dx} = \\mathbb{1}(|x| - t > 0)\n",
"$$"
]
},
{
"cell_type": "code",
"execution_count": 17,
"metadata": {
"ExecuteTime": {
"end_time": "2018-12-03T16:52:40.420206Z",
"start_time": "2018-12-03T16:52:40.395286Z"
}
},
"outputs": [
{
"data": {
"application/javascript": [
"/* Put everything inside the global mpl namespace */\n",
"window.mpl = {};\n",
"\n",
"\n",
"mpl.get_websocket_type = function() {\n",
" if (typeof(WebSocket) !== 'undefined') {\n",
" return WebSocket;\n",
" } else if (typeof(MozWebSocket) !== 'undefined') {\n",
" return MozWebSocket;\n",
" } else {\n",
" alert('Your browser does not have WebSocket support.' +\n",
" 'Please try Chrome, Safari or Firefox ≥ 6. ' +\n",
" 'Firefox 4 and 5 are also supported but you ' +\n",
" 'have to enable WebSockets in about:config.');\n",
" };\n",
"}\n",
"\n",
"mpl.figure = function(figure_id, websocket, ondownload, parent_element) {\n",
" this.id = figure_id;\n",
"\n",
" this.ws = websocket;\n",
"\n",
" this.supports_binary = (this.ws.binaryType != undefined);\n",
"\n",
" if (!this.supports_binary) {\n",
" var warnings = document.getElementById(\"mpl-warnings\");\n",
" if (warnings) {\n",
" warnings.style.display = 'block';\n",
" warnings.textContent = (\n",
" \"This browser does not support binary websocket messages. \" +\n",
" \"Performance may be slow.\");\n",
" }\n",
" }\n",
"\n",
" this.imageObj = new Image();\n",
"\n",
" this.context = undefined;\n",
" this.message = undefined;\n",
" this.canvas = undefined;\n",
" this.rubberband_canvas = undefined;\n",
" this.rubberband_context = undefined;\n",
" this.format_dropdown = undefined;\n",
"\n",
" this.image_mode = 'full';\n",
"\n",
" this.root = $('');\n",
" this._root_extra_style(this.root)\n",
" this.root.attr('style', 'display: inline-block');\n",
"\n",
" $(parent_element).append(this.root);\n",
"\n",
" this._init_header(this);\n",
" this._init_canvas(this);\n",
" this._init_toolbar(this);\n",
"\n",
" var fig = this;\n",
"\n",
" this.waiting = false;\n",
"\n",
" this.ws.onopen = function () {\n",
" fig.send_message(\"supports_binary\", {value: fig.supports_binary});\n",
" fig.send_message(\"send_image_mode\", {});\n",
" if (mpl.ratio != 1) {\n",
" fig.send_message(\"set_dpi_ratio\", {'dpi_ratio': mpl.ratio});\n",
" }\n",
" fig.send_message(\"refresh\", {});\n",
" }\n",
"\n",
" this.imageObj.onload = function() {\n",
" if (fig.image_mode == 'full') {\n",
" // Full images could contain transparency (where diff images\n",
" // almost always do), so we need to clear the canvas so that\n",
" // there is no ghosting.\n",
" fig.context.clearRect(0, 0, fig.canvas.width, fig.canvas.height);\n",
" }\n",
" fig.context.drawImage(fig.imageObj, 0, 0);\n",
" };\n",
"\n",
" this.imageObj.onunload = function() {\n",
" fig.ws.close();\n",
" }\n",
"\n",
" this.ws.onmessage = this._make_on_message_function(this);\n",
"\n",
" this.ondownload = ondownload;\n",
"}\n",
"\n",
"mpl.figure.prototype._init_header = function() {\n",
" var titlebar = $(\n",
" '');\n",
" var titletext = $(\n",
" '');\n",
" titlebar.append(titletext)\n",
" this.root.append(titlebar);\n",
" this.header = titletext[0];\n",
"}\n",
"\n",
"\n",
"\n",
"mpl.figure.prototype._canvas_extra_style = function(canvas_div) {\n",
"\n",
"}\n",
"\n",
"\n",
"mpl.figure.prototype._root_extra_style = function(canvas_div) {\n",
"\n",
"}\n",
"\n",
"mpl.figure.prototype._init_canvas = function() {\n",
" var fig = this;\n",
"\n",
" var canvas_div = $('');\n",
"\n",
" canvas_div.attr('style', 'position: relative; clear: both; outline: 0');\n",
"\n",
" function canvas_keyboard_event(event) {\n",
" return fig.key_event(event, event['data']);\n",
" }\n",
"\n",
" canvas_div.keydown('key_press', canvas_keyboard_event);\n",
" canvas_div.keyup('key_release', canvas_keyboard_event);\n",
" this.canvas_div = canvas_div\n",
" this._canvas_extra_style(canvas_div)\n",
" this.root.append(canvas_div);\n",
"\n",
" var canvas = $('');\n",
" canvas.addClass('mpl-canvas');\n",
" canvas.attr('style', \"left: 0; top: 0; z-index: 0; outline: 0\")\n",
"\n",
" this.canvas = canvas[0];\n",
" this.context = canvas[0].getContext(\"2d\");\n",
"\n",
" var backingStore = this.context.backingStorePixelRatio ||\n",
"\tthis.context.webkitBackingStorePixelRatio ||\n",
"\tthis.context.mozBackingStorePixelRatio ||\n",
"\tthis.context.msBackingStorePixelRatio ||\n",
"\tthis.context.oBackingStorePixelRatio ||\n",
"\tthis.context.backingStorePixelRatio || 1;\n",
"\n",
" mpl.ratio = (window.devicePixelRatio || 1) / backingStore;\n",
"\n",
" var rubberband = $('');\n",
" rubberband.attr('style', \"position: absolute; left: 0; top: 0; z-index: 1;\")\n",
"\n",
" var pass_mouse_events = true;\n",
"\n",
" canvas_div.resizable({\n",
" start: function(event, ui) {\n",
" pass_mouse_events = false;\n",
" },\n",
" resize: function(event, ui) {\n",
" fig.request_resize(ui.size.width, ui.size.height);\n",
" },\n",
" stop: function(event, ui) {\n",
" pass_mouse_events = true;\n",
" fig.request_resize(ui.size.width, ui.size.height);\n",
" },\n",
" });\n",
"\n",
" function mouse_event_fn(event) {\n",
" if (pass_mouse_events)\n",
" return fig.mouse_event(event, event['data']);\n",
" }\n",
"\n",
" rubberband.mousedown('button_press', mouse_event_fn);\n",
" rubberband.mouseup('button_release', mouse_event_fn);\n",
" // Throttle sequential mouse events to 1 every 20ms.\n",
" rubberband.mousemove('motion_notify', mouse_event_fn);\n",
"\n",
" rubberband.mouseenter('figure_enter', mouse_event_fn);\n",
" rubberband.mouseleave('figure_leave', mouse_event_fn);\n",
"\n",
" canvas_div.on(\"wheel\", function (event) {\n",
" event = event.originalEvent;\n",
" event['data'] = 'scroll'\n",
" if (event.deltaY < 0) {\n",
" event.step = 1;\n",
" } else {\n",
" event.step = -1;\n",
" }\n",
" mouse_event_fn(event);\n",
" });\n",
"\n",
" canvas_div.append(canvas);\n",
" canvas_div.append(rubberband);\n",
"\n",
" this.rubberband = rubberband;\n",
" this.rubberband_canvas = rubberband[0];\n",
" this.rubberband_context = rubberband[0].getContext(\"2d\");\n",
" this.rubberband_context.strokeStyle = \"#000000\";\n",
"\n",
" this._resize_canvas = function(width, height) {\n",
" // Keep the size of the canvas, canvas container, and rubber band\n",
" // canvas in synch.\n",
" canvas_div.css('width', width)\n",
" canvas_div.css('height', height)\n",
"\n",
" canvas.attr('width', width * mpl.ratio);\n",
" canvas.attr('height', height * mpl.ratio);\n",
" canvas.attr('style', 'width: ' + width + 'px; height: ' + height + 'px;');\n",
"\n",
" rubberband.attr('width', width);\n",
" rubberband.attr('height', height);\n",
" }\n",
"\n",
" // Set the figure to an initial 600x600px, this will subsequently be updated\n",
" // upon first draw.\n",
" this._resize_canvas(600, 600);\n",
"\n",
" // Disable right mouse context menu.\n",
" $(this.rubberband_canvas).bind(\"contextmenu\",function(e){\n",
" return false;\n",
" });\n",
"\n",
" function set_focus () {\n",
" canvas.focus();\n",
" canvas_div.focus();\n",
" }\n",
"\n",
" window.setTimeout(set_focus, 100);\n",
"}\n",
"\n",
"mpl.figure.prototype._init_toolbar = function() {\n",
" var fig = this;\n",
"\n",
" var nav_element = $('')\n",
" nav_element.attr('style', 'width: 100%');\n",
" this.root.append(nav_element);\n",
"\n",
" // Define a callback function for later on.\n",
" function toolbar_event(event) {\n",
" return fig.toolbar_button_onclick(event['data']);\n",
" }\n",
" function toolbar_mouse_event(event) {\n",
" return fig.toolbar_button_onmouseover(event['data']);\n",
" }\n",
"\n",
" for(var toolbar_ind in mpl.toolbar_items) {\n",
" var name = mpl.toolbar_items[toolbar_ind][0];\n",
" var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
" var image = mpl.toolbar_items[toolbar_ind][2];\n",
" var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
"\n",
" if (!name) {\n",
" // put a spacer in here.\n",
" continue;\n",
" }\n",
" var button = $('');\n",
" button.addClass('ui-button ui-widget ui-state-default ui-corner-all ' +\n",
" 'ui-button-icon-only');\n",
" button.attr('role', 'button');\n",
" button.attr('aria-disabled', 'false');\n",
" button.click(method_name, toolbar_event);\n",
" button.mouseover(tooltip, toolbar_mouse_event);\n",
"\n",
" var icon_img = $('');\n",
" icon_img.addClass('ui-button-icon-primary ui-icon');\n",
" icon_img.addClass(image);\n",
" icon_img.addClass('ui-corner-all');\n",
"\n",
" var tooltip_span = $('');\n",
" tooltip_span.addClass('ui-button-text');\n",
" tooltip_span.html(tooltip);\n",
"\n",
" button.append(icon_img);\n",
" button.append(tooltip_span);\n",
"\n",
" nav_element.append(button);\n",
" }\n",
"\n",
" var fmt_picker_span = $('');\n",
"\n",
" var fmt_picker = $('');\n",
" fmt_picker.addClass('mpl-toolbar-option ui-widget ui-widget-content');\n",
" fmt_picker_span.append(fmt_picker);\n",
" nav_element.append(fmt_picker_span);\n",
" this.format_dropdown = fmt_picker[0];\n",
"\n",
" for (var ind in mpl.extensions) {\n",
" var fmt = mpl.extensions[ind];\n",
" var option = $(\n",
" '', {selected: fmt === mpl.default_extension}).html(fmt);\n",
" fmt_picker.append(option)\n",
" }\n",
"\n",
" // Add hover states to the ui-buttons\n",
" $( \".ui-button\" ).hover(\n",
" function() { $(this).addClass(\"ui-state-hover\");},\n",
" function() { $(this).removeClass(\"ui-state-hover\");}\n",
" );\n",
"\n",
" var status_bar = $('');\n",
" nav_element.append(status_bar);\n",
" this.message = status_bar[0];\n",
"}\n",
"\n",
"mpl.figure.prototype.request_resize = function(x_pixels, y_pixels) {\n",
" // Request matplotlib to resize the figure. Matplotlib will then trigger a resize in the client,\n",
" // which will in turn request a refresh of the image.\n",
" this.send_message('resize', {'width': x_pixels, 'height': y_pixels});\n",
"}\n",
"\n",
"mpl.figure.prototype.send_message = function(type, properties) {\n",
" properties['type'] = type;\n",
" properties['figure_id'] = this.id;\n",
" this.ws.send(JSON.stringify(properties));\n",
"}\n",
"\n",
"mpl.figure.prototype.send_draw_message = function() {\n",
" if (!this.waiting) {\n",
" this.waiting = true;\n",
" this.ws.send(JSON.stringify({type: \"draw\", figure_id: this.id}));\n",
" }\n",
"}\n",
"\n",
"\n",
"mpl.figure.prototype.handle_save = function(fig, msg) {\n",
" var format_dropdown = fig.format_dropdown;\n",
" var format = format_dropdown.options[format_dropdown.selectedIndex].value;\n",
" fig.ondownload(fig, format);\n",
"}\n",
"\n",
"\n",
"mpl.figure.prototype.handle_resize = function(fig, msg) {\n",
" var size = msg['size'];\n",
" if (size[0] != fig.canvas.width || size[1] != fig.canvas.height) {\n",
" fig._resize_canvas(size[0], size[1]);\n",
" fig.send_message(\"refresh\", {});\n",
" };\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_rubberband = function(fig, msg) {\n",
" var x0 = msg['x0'] / mpl.ratio;\n",
" var y0 = (fig.canvas.height - msg['y0']) / mpl.ratio;\n",
" var x1 = msg['x1'] / mpl.ratio;\n",
" var y1 = (fig.canvas.height - msg['y1']) / mpl.ratio;\n",
" x0 = Math.floor(x0) + 0.5;\n",
" y0 = Math.floor(y0) + 0.5;\n",
" x1 = Math.floor(x1) + 0.5;\n",
" y1 = Math.floor(y1) + 0.5;\n",
" var min_x = Math.min(x0, x1);\n",
" var min_y = Math.min(y0, y1);\n",
" var width = Math.abs(x1 - x0);\n",
" var height = Math.abs(y1 - y0);\n",
"\n",
" fig.rubberband_context.clearRect(\n",
" 0, 0, fig.canvas.width, fig.canvas.height);\n",
"\n",
" fig.rubberband_context.strokeRect(min_x, min_y, width, height);\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_figure_label = function(fig, msg) {\n",
" // Updates the figure title.\n",
" fig.header.textContent = msg['label'];\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_cursor = function(fig, msg) {\n",
" var cursor = msg['cursor'];\n",
" switch(cursor)\n",
" {\n",
" case 0:\n",
" cursor = 'pointer';\n",
" break;\n",
" case 1:\n",
" cursor = 'default';\n",
" break;\n",
" case 2:\n",
" cursor = 'crosshair';\n",
" break;\n",
" case 3:\n",
" cursor = 'move';\n",
" break;\n",
" }\n",
" fig.rubberband_canvas.style.cursor = cursor;\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_message = function(fig, msg) {\n",
" fig.message.textContent = msg['message'];\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_draw = function(fig, msg) {\n",
" // Request the server to send over a new figure.\n",
" fig.send_draw_message();\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_image_mode = function(fig, msg) {\n",
" fig.image_mode = msg['mode'];\n",
"}\n",
"\n",
"mpl.figure.prototype.updated_canvas_event = function() {\n",
" // Called whenever the canvas gets updated.\n",
" this.send_message(\"ack\", {});\n",
"}\n",
"\n",
"// A function to construct a web socket function for onmessage handling.\n",
"// Called in the figure constructor.\n",
"mpl.figure.prototype._make_on_message_function = function(fig) {\n",
" return function socket_on_message(evt) {\n",
" if (evt.data instanceof Blob) {\n",
" /* FIXME: We get \"Resource interpreted as Image but\n",
" * transferred with MIME type text/plain:\" errors on\n",
" * Chrome. But how to set the MIME type? It doesn't seem\n",
" * to be part of the websocket stream */\n",
" evt.data.type = \"image/png\";\n",
"\n",
" /* Free the memory for the previous frames */\n",
" if (fig.imageObj.src) {\n",
" (window.URL || window.webkitURL).revokeObjectURL(\n",
" fig.imageObj.src);\n",
" }\n",
"\n",
" fig.imageObj.src = (window.URL || window.webkitURL).createObjectURL(\n",
" evt.data);\n",
" fig.updated_canvas_event();\n",
" fig.waiting = false;\n",
" return;\n",
" }\n",
" else if (typeof evt.data === 'string' && evt.data.slice(0, 21) == \"data:image/png;base64\") {\n",
" fig.imageObj.src = evt.data;\n",
" fig.updated_canvas_event();\n",
" fig.waiting = false;\n",
" return;\n",
" }\n",
"\n",
" var msg = JSON.parse(evt.data);\n",
" var msg_type = msg['type'];\n",
"\n",
" // Call the \"handle_{type}\" callback, which takes\n",
" // the figure and JSON message as its only arguments.\n",
" try {\n",
" var callback = fig[\"handle_\" + msg_type];\n",
" } catch (e) {\n",
" console.log(\"No handler for the '\" + msg_type + \"' message type: \", msg);\n",
" return;\n",
" }\n",
"\n",
" if (callback) {\n",
" try {\n",
" // console.log(\"Handling '\" + msg_type + \"' message: \", msg);\n",
" callback(fig, msg);\n",
" } catch (e) {\n",
" console.log(\"Exception inside the 'handler_\" + msg_type + \"' callback:\", e, e.stack, msg);\n",
" }\n",
" }\n",
" };\n",
"}\n",
"\n",
"// from http://stackoverflow.com/questions/1114465/getting-mouse-location-in-canvas\n",
"mpl.findpos = function(e) {\n",
" //this section is from http://www.quirksmode.org/js/events_properties.html\n",
" var targ;\n",
" if (!e)\n",
" e = window.event;\n",
" if (e.target)\n",
" targ = e.target;\n",
" else if (e.srcElement)\n",
" targ = e.srcElement;\n",
" if (targ.nodeType == 3) // defeat Safari bug\n",
" targ = targ.parentNode;\n",
"\n",
" // jQuery normalizes the pageX and pageY\n",
" // pageX,Y are the mouse positions relative to the document\n",
" // offset() returns the position of the element relative to the document\n",
" var x = e.pageX - $(targ).offset().left;\n",
" var y = e.pageY - $(targ).offset().top;\n",
"\n",
" return {\"x\": x, \"y\": y};\n",
"};\n",
"\n",
"/*\n",
" * return a copy of an object with only non-object keys\n",
" * we need this to avoid circular references\n",
" * http://stackoverflow.com/a/24161582/3208463\n",
" */\n",
"function simpleKeys (original) {\n",
" return Object.keys(original).reduce(function (obj, key) {\n",
" if (typeof original[key] !== 'object')\n",
" obj[key] = original[key]\n",
" return obj;\n",
" }, {});\n",
"}\n",
"\n",
"mpl.figure.prototype.mouse_event = function(event, name) {\n",
" var canvas_pos = mpl.findpos(event)\n",
"\n",
" if (name === 'button_press')\n",
" {\n",
" this.canvas.focus();\n",
" this.canvas_div.focus();\n",
" }\n",
"\n",
" var x = canvas_pos.x * mpl.ratio;\n",
" var y = canvas_pos.y * mpl.ratio;\n",
"\n",
" this.send_message(name, {x: x, y: y, button: event.button,\n",
" step: event.step,\n",
" guiEvent: simpleKeys(event)});\n",
"\n",
" /* This prevents the web browser from automatically changing to\n",
" * the text insertion cursor when the button is pressed. We want\n",
" * to control all of the cursor setting manually through the\n",
" * 'cursor' event from matplotlib */\n",
" event.preventDefault();\n",
" return false;\n",
"}\n",
"\n",
"mpl.figure.prototype._key_event_extra = function(event, name) {\n",
" // Handle any extra behaviour associated with a key event\n",
"}\n",
"\n",
"mpl.figure.prototype.key_event = function(event, name) {\n",
"\n",
" // Prevent repeat events\n",
" if (name == 'key_press')\n",
" {\n",
" if (event.which === this._key)\n",
" return;\n",
" else\n",
" this._key = event.which;\n",
" }\n",
" if (name == 'key_release')\n",
" this._key = null;\n",
"\n",
" var value = '';\n",
" if (event.ctrlKey && event.which != 17)\n",
" value += \"ctrl+\";\n",
" if (event.altKey && event.which != 18)\n",
" value += \"alt+\";\n",
" if (event.shiftKey && event.which != 16)\n",
" value += \"shift+\";\n",
"\n",
" value += 'k';\n",
" value += event.which.toString();\n",
"\n",
" this._key_event_extra(event, name);\n",
"\n",
" this.send_message(name, {key: value,\n",
" guiEvent: simpleKeys(event)});\n",
" return false;\n",
"}\n",
"\n",
"mpl.figure.prototype.toolbar_button_onclick = function(name) {\n",
" if (name == 'download') {\n",
" this.handle_save(this, null);\n",
" } else {\n",
" this.send_message(\"toolbar_button\", {name: name});\n",
" }\n",
"};\n",
"\n",
"mpl.figure.prototype.toolbar_button_onmouseover = function(tooltip) {\n",
" this.message.textContent = tooltip;\n",
"};\n",
"mpl.toolbar_items = [[\"Home\", \"Reset original view\", \"fa fa-home icon-home\", \"home\"], [\"Back\", \"Back to previous view\", \"fa fa-arrow-left icon-arrow-left\", \"back\"], [\"Forward\", \"Forward to next view\", \"fa fa-arrow-right icon-arrow-right\", \"forward\"], [\"\", \"\", \"\", \"\"], [\"Pan\", \"Pan axes with left mouse, zoom with right\", \"fa fa-arrows icon-move\", \"pan\"], [\"Zoom\", \"Zoom to rectangle\", \"fa fa-square-o icon-check-empty\", \"zoom\"], [\"\", \"\", \"\", \"\"], [\"Download\", \"Download plot\", \"fa fa-floppy-o icon-save\", \"download\"]];\n",
"\n",
"mpl.extensions = [\"eps\", \"jpeg\", \"pdf\", \"png\", \"ps\", \"raw\", \"svg\", \"tif\"];\n",
"\n",
"mpl.default_extension = \"png\";var comm_websocket_adapter = function(comm) {\n",
" // Create a \"websocket\"-like object which calls the given IPython comm\n",
" // object with the appropriate methods. Currently this is a non binary\n",
" // socket, so there is still some room for performance tuning.\n",
" var ws = {};\n",
"\n",
" ws.close = function() {\n",
" comm.close()\n",
" };\n",
" ws.send = function(m) {\n",
" //console.log('sending', m);\n",
" comm.send(m);\n",
" };\n",
" // Register the callback with on_msg.\n",
" comm.on_msg(function(msg) {\n",
" //console.log('receiving', msg['content']['data'], msg);\n",
" // Pass the mpl event to the overridden (by mpl) onmessage function.\n",
" ws.onmessage(msg['content']['data'])\n",
" });\n",
" return ws;\n",
"}\n",
"\n",
"mpl.mpl_figure_comm = function(comm, msg) {\n",
" // This is the function which gets called when the mpl process\n",
" // starts-up an IPython Comm through the \"matplotlib\" channel.\n",
"\n",
" var id = msg.content.data.id;\n",
" // Get hold of the div created by the display call when the Comm\n",
" // socket was opened in Python.\n",
" var element = $(\"#\" + id);\n",
" var ws_proxy = comm_websocket_adapter(comm)\n",
"\n",
" function ondownload(figure, format) {\n",
" window.open(figure.imageObj.src);\n",
" }\n",
"\n",
" var fig = new mpl.figure(id, ws_proxy,\n",
" ondownload,\n",
" element.get(0));\n",
"\n",
" // Call onopen now - mpl needs it, as it is assuming we've passed it a real\n",
" // web socket which is closed, not our websocket->open comm proxy.\n",
" ws_proxy.onopen();\n",
"\n",
" fig.parent_element = element.get(0);\n",
" fig.cell_info = mpl.find_output_cell(\"\");\n",
" if (!fig.cell_info) {\n",
" console.error(\"Failed to find cell for figure\", id, fig);\n",
" return;\n",
" }\n",
"\n",
" var output_index = fig.cell_info[2]\n",
" var cell = fig.cell_info[0];\n",
"\n",
"};\n",
"\n",
"mpl.figure.prototype.handle_close = function(fig, msg) {\n",
" var width = fig.canvas.width/mpl.ratio\n",
" fig.root.unbind('remove')\n",
"\n",
" // Update the output cell to use the data from the current canvas.\n",
" fig.push_to_output();\n",
" var dataURL = fig.canvas.toDataURL();\n",
" // Re-enable the keyboard manager in IPython - without this line, in FF,\n",
" // the notebook keyboard shortcuts fail.\n",
" IPython.keyboard_manager.enable()\n",
" $(fig.parent_element).html('');\n",
" fig.close_ws(fig, msg);\n",
"}\n",
"\n",
"mpl.figure.prototype.close_ws = function(fig, msg){\n",
" fig.send_message('closing', msg);\n",
" // fig.ws.close()\n",
"}\n",
"\n",
"mpl.figure.prototype.push_to_output = function(remove_interactive) {\n",
" // Turn the data on the canvas into data in the output cell.\n",
" var width = this.canvas.width/mpl.ratio\n",
" var dataURL = this.canvas.toDataURL();\n",
" this.cell_info[1]['text/html'] = '';\n",
"}\n",
"\n",
"mpl.figure.prototype.updated_canvas_event = function() {\n",
" // Tell IPython that the notebook contents must change.\n",
" IPython.notebook.set_dirty(true);\n",
" this.send_message(\"ack\", {});\n",
" var fig = this;\n",
" // Wait a second, then push the new image to the DOM so\n",
" // that it is saved nicely (might be nice to debounce this).\n",
" setTimeout(function () { fig.push_to_output() }, 1000);\n",
"}\n",
"\n",
"mpl.figure.prototype._init_toolbar = function() {\n",
" var fig = this;\n",
"\n",
" var nav_element = $('')\n",
" nav_element.attr('style', 'width: 100%');\n",
" this.root.append(nav_element);\n",
"\n",
" // Define a callback function for later on.\n",
" function toolbar_event(event) {\n",
" return fig.toolbar_button_onclick(event['data']);\n",
" }\n",
" function toolbar_mouse_event(event) {\n",
" return fig.toolbar_button_onmouseover(event['data']);\n",
" }\n",
"\n",
" for(var toolbar_ind in mpl.toolbar_items){\n",
" var name = mpl.toolbar_items[toolbar_ind][0];\n",
" var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
" var image = mpl.toolbar_items[toolbar_ind][2];\n",
" var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
"\n",
" if (!name) { continue; };\n",
"\n",
" var button = $('');\n",
" button.click(method_name, toolbar_event);\n",
" button.mouseover(tooltip, toolbar_mouse_event);\n",
" nav_element.append(button);\n",
" }\n",
"\n",
" // Add the status bar.\n",
" var status_bar = $('');\n",
" nav_element.append(status_bar);\n",
" this.message = status_bar[0];\n",
"\n",
" // Add the close button to the window.\n",
" var buttongrp = $('');\n",
" var button = $('');\n",
" button.click(function (evt) { fig.handle_close(fig, {}); } );\n",
" button.mouseover('Stop Interaction', toolbar_mouse_event);\n",
" buttongrp.append(button);\n",
" var titlebar = this.root.find($('.ui-dialog-titlebar'));\n",
" titlebar.prepend(buttongrp);\n",
"}\n",
"\n",
"mpl.figure.prototype._root_extra_style = function(el){\n",
" var fig = this\n",
" el.on(\"remove\", function(){\n",
"\tfig.close_ws(fig, {});\n",
" });\n",
"}\n",
"\n",
"mpl.figure.prototype._canvas_extra_style = function(el){\n",
" // this is important to make the div 'focusable\n",
" el.attr('tabindex', 0)\n",
" // reach out to IPython and tell the keyboard manager to turn it's self\n",
" // off when our div gets focus\n",
"\n",
" // location in version 3\n",
" if (IPython.notebook.keyboard_manager) {\n",
" IPython.notebook.keyboard_manager.register_events(el);\n",
" }\n",
" else {\n",
" // location in version 2\n",
" IPython.keyboard_manager.register_events(el);\n",
" }\n",
"\n",
"}\n",
"\n",
"mpl.figure.prototype._key_event_extra = function(event, name) {\n",
" var manager = IPython.notebook.keyboard_manager;\n",
" if (!manager)\n",
" manager = IPython.keyboard_manager;\n",
"\n",
" // Check for shift+enter\n",
" if (event.shiftKey && event.which == 13) {\n",
" this.canvas_div.blur();\n",
" event.shiftKey = false;\n",
" // Send a \"J\" for go to next cell\n",
" event.which = 74;\n",
" event.keyCode = 74;\n",
" manager.command_mode();\n",
" manager.handle_keydown(event);\n",
" }\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_save = function(fig, msg) {\n",
" fig.ondownload(fig, null);\n",
"}\n",
"\n",
"\n",
"mpl.find_output_cell = function(html_output) {\n",
" // Return the cell and output element which can be found *uniquely* in the notebook.\n",
" // Note - this is a bit hacky, but it is done because the \"notebook_saving.Notebook\"\n",
" // IPython event is triggered only after the cells have been serialised, which for\n",
" // our purposes (turning an active figure into a static one), is too late.\n",
" var cells = IPython.notebook.get_cells();\n",
" var ncells = cells.length;\n",
" for (var i=0; i= 3 moved mimebundle to data attribute of output\n",
" data = data.data;\n",
" }\n",
" if (data['text/html'] == html_output) {\n",
" return [cell, data, j];\n",
" }\n",
" }\n",
" }\n",
" }\n",
"}\n",
"\n",
"// Register the function which deals with the matplotlib target/channel.\n",
"// The kernel may be null if the page has been refreshed.\n",
"if (IPython.notebook.kernel != null) {\n",
" IPython.notebook.kernel.comm_manager.register_target('matplotlib', mpl.mpl_figure_comm);\n",
"}\n"
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/html": [
""
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"plt.figure()\n",
"t = 1.6\n",
"x = np.linspace(-5,5,100)\n",
"w = shrink(x, t)\n",
"dwdt = -np.sign(w)\n",
"plt.plot(x, w, label='w')\n",
"plt.plot(x, dwdt, label='dw/dt')\n",
"plt.legend(frameon=True)\n",
"plt.axhline(y=0, color='k', linewidth=.5)\n",
"plt.axvline(x=0, color='k', linewidth=.5)\n",
"plt.grid(ls='dashed')"
]
},
{
"cell_type": "code",
"execution_count": 18,
"metadata": {
"ExecuteTime": {
"end_time": "2018-12-03T16:52:41.848811Z",
"start_time": "2018-12-03T16:52:40.421386Z"
}
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Gradient at T=5 is -0.0592\n",
"Gradient at T=20 is 0.0342\n"
]
}
],
"source": [
"def dLdX_hat(x_hat, im):\n",
" return 2*(x_hat-im)/x_hat.size\n",
"\n",
"def dX_hatdw(x_hat):\n",
" coeffs = pywt.wavedec2(x_hat, wave, axes=(0,1), level=J)\n",
" return coeffs[1][0][:,:,1]\n",
"\n",
"def dwdT(w, T):\n",
" coeffs = pywt.wavedec2(im_noise, wave, axes=(0,1), level=J)\n",
" g = shrink(coeffs[1][0][:,:,1], T)\n",
" g = -np.sign(g)\n",
" return np.sum(g*w)\n",
"\n",
"def np_backprop(T):\n",
" return dwdT(dX_hatdw(dLdX_hat(F(T), im2)), T)\n",
"\n",
"print('Gradient at T=5 is {:.4f}'.format(np_backprop(5)))\n",
"print('Gradient at T=20 is {:.4f}'.format(np_backprop(20)))"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## autograd (using pytorch)\n",
"Pytorch does have its own soft thresh function, but it only gives gradients w.r.t the input.\n",
"\n",
"To work around this, we have to define our own autograd function for the soft threshold which gives gradients w.r.t. the input and the thresholds:"
]
},
{
"cell_type": "code",
"execution_count": 19,
"metadata": {
"ExecuteTime": {
"end_time": "2018-12-03T16:52:41.907756Z",
"start_time": "2018-12-03T16:52:41.849902Z"
}
},
"outputs": [
{
"data": {
"text/plain": [
"True"
]
},
"execution_count": 19,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"from torch.autograd import Function\n",
"class SoftShrink_fn(Function):\n",
" @staticmethod\n",
" def forward(ctx, x, t):\n",
" y = nn.functional.softshrink(x, t.item())\n",
" ctx.save_for_backward(torch.sign(y))\n",
" return y\n",
"\n",
" @staticmethod\n",
" def backward(ctx, dy):\n",
" din, = ctx.saved_tensors\n",
" dx, dt = None, None\n",
" if ctx.needs_input_grad[0]:\n",
" dx = dy * din * din\n",
" if ctx.needs_input_grad[1]:\n",
" dt = -torch.sum(dy * din)\n",
" return dx, dt\n",
"\n",
"class SoftShrink(nn.Module): \n",
" def __init__(self, t_init): \n",
" super().__init__() \n",
" self.t = nn.Parameter(torch.tensor(t_init)) \n",
" \n",
" def forward(self, x): \n",
" \"\"\" Applies Soft Thresholding to x \"\"\" \n",
" return SoftShrink_fn.apply(x, self.t)\n",
" \n",
"from torch.autograd import gradcheck\n",
"x = torch.randn(10,10, requires_grad=True, dtype=torch.double)\n",
"t = torch.tensor(1., requires_grad=True, dtype=torch.double)\n",
"gradcheck(SoftShrink_fn.apply, (x,t))"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Use this to calculate the gradient at $T=5$ and $T=20$ as we did for the finite differences above"
]
},
{
"cell_type": "code",
"execution_count": 21,
"metadata": {
"ExecuteTime": {
"end_time": "2018-12-03T16:53:27.431062Z",
"start_time": "2018-12-03T16:53:26.567371Z"
}
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Gradient at T=5 is -0.0592\n",
"Gradient at T=20 is 0.0342\n"
]
}
],
"source": [
"dwt = DWTForward(C=3,J=3, wave='db1')\n",
"iwt = DWTInverse(C=3, wave='db1')\n",
"X = torch.tensor(im.astype('float32').transpose((2,0,1))[None,:])\n",
"Y = torch.tensor(im_noise.astype('float32').transpose((2,0,1))[None,:], \n",
" requires_grad=True)\n",
"\n",
"# Calculate the mean removed image\n",
"coeffs = dwt(Y)\n",
"zeros = dwt(torch.zeros_like(Y))\n",
"zeros[0] = coeffs[0]\n",
"X2 = X - iwt(zeros)\n",
"shrinker = SoftShrink(5.0)\n",
"\n",
"def F_torch(T):\n",
" # Shrink the LH of the first scale\n",
" coeffs = dwt(Y)\n",
" zeros = dwt(torch.zeros_like(Y))\n",
" shrinker.t.data = torch.tensor(T)\n",
" c = shrinker(coeffs[1][-1])\n",
" c[:,0] = 0.\n",
" c[:,2] = 0.\n",
" c[:,1,1:] = 0.\n",
" coeffs_shrink = [\n",
" torch.zeros_like(coeffs[0]),\n",
" [torch.zeros_like(coeffs[1][0]),\n",
" torch.zeros_like(coeffs[1][1]),\n",
" c\n",
" ]\n",
" ]\n",
" return iwt(coeffs_shrink)\n",
"\n",
"def L_torch(T):\n",
" return torch.mean((X2 - F_torch(T))**2)\n",
"\n",
"y = L_torch(5.0)\n",
"y.backward(retain_graph=True)\n",
"print('Gradient at T=5 is {:.4f}'.format(shrinker.t.grad.item()))\n",
"shrinker.t.grad.zero_()\n",
"y = L_torch(20.0)\n",
"y.backward(retain_graph=True)\n",
"print('Gradient at T=20 is {:.4f}'.format(shrinker.t.grad.item()))"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Good! this agrees with the finite differences approximation to the gradient. Now that we have a function to calculate gradients, let us minimize!"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# Optimize\n",
"We will need to slightly modify the above softshrink function to work with channels of input:"
]
},
{
"cell_type": "code",
"execution_count": 23,
"metadata": {
"ExecuteTime": {
"end_time": "2018-12-03T16:53:47.648462Z",
"start_time": "2018-12-03T16:53:47.554184Z"
}
},
"outputs": [
{
"data": {
"text/plain": [
"True"
]
},
"execution_count": 23,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"class SoftShrink_ch_fn(Function):\n",
" @staticmethod\n",
" def forward(ctx, x, t, C):\n",
" y = torch.zeros_like(x)\n",
" for c in range(C):\n",
" for i in range(3):\n",
" y[:,c,i] = nn.functional.softshrink(x[:,c,i], t.data[c,i])\n",
" ctx.save_for_backward(torch.sign(y))\n",
" return y\n",
"\n",
" @staticmethod\n",
" def backward(ctx, dy):\n",
" din, = ctx.saved_tensors\n",
" dx, dt = None, None\n",
" if ctx.needs_input_grad[0]:\n",
" dx = dy * din * din\n",
" if ctx.needs_input_grad[1]:\n",
" dt = -torch.sum(dy * din, dim=(0,3,4))\n",
" return dx, dt, None\n",
"\n",
"class SoftShrink_ch(nn.Module): \n",
" def __init__(self, C, t_init, t_grad=True):\n",
" super().__init__()\n",
" assert t_init.shape[-1] == 3\n",
" assert t_init.shape[0] == C\n",
" self.t = nn.Parameter(torch.tensor(t_init).float())\n",
" self.constrain = nn.ReLU()\n",
" self.C = C\n",
"\n",
" @property\n",
" def thresh(self):\n",
" return self.constrain(self.t)\n",
"\n",
" def forward(self, x):\n",
" \"\"\" Applies Soft Thresholding to x \"\"\"\n",
" if x.shape == torch.Size([0]):\n",
" return x\n",
" else:\n",
" assert x.shape[1] == self.C\n",
" return SoftShrink_ch_fn.apply(x, self.thresh, self.C)\n",
" \n",
"x = torch.randn(1,3,3,4,4, dtype=torch.double, requires_grad=True)\n",
"t = torch.rand(3,3, dtype=torch.double, requires_grad=True)\n",
"y = SoftShrink_ch_fn.apply(x, t, 3)\n",
"gradcheck(SoftShrink_ch_fn.apply, (x,t,3))"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Now that we have this, create the 'net' that shrinks coefficients:"
]
},
{
"cell_type": "code",
"execution_count": 27,
"metadata": {
"ExecuteTime": {
"end_time": "2018-12-03T16:54:07.080720Z",
"start_time": "2018-12-03T16:54:07.073215Z"
}
},
"outputs": [],
"source": [
"class MyThresh(nn.Module):\n",
" def __init__(self, t_init=None):\n",
" super().__init__()\n",
" if t_init is None:\n",
" t_init = np.random.rand(3,3,3)\n",
" self.dwt = DWTForward(C=3,J=3, wave=wave)\n",
" self.iwt = DWTInverse(C=3, wave=wave)\n",
" self.shrinkers = nn.ModuleList([\n",
" SoftShrink_ch(C=3, t_init=t_init[0]),\n",
" SoftShrink_ch(C=3, t_init=t_init[1]),\n",
" SoftShrink_ch(C=3, t_init=t_init[2])\n",
" ])\n",
" \n",
" def forward(self, x):\n",
" coeffs = self.dwt(x)\n",
" coeffs[1][0] = self.shrinkers[0](coeffs[1][0])\n",
" coeffs[1][1] = self.shrinkers[1](coeffs[1][1])\n",
" coeffs[1][2] = self.shrinkers[2](coeffs[1][2])\n",
" y = self.iwt(coeffs)\n",
" return y"
]
},
{
"cell_type": "code",
"execution_count": 28,
"metadata": {
"ExecuteTime": {
"end_time": "2018-12-03T16:54:07.176537Z",
"start_time": "2018-12-03T16:54:07.082613Z"
},
"scrolled": true
},
"outputs": [],
"source": [
"def get_lr(optim):\n",
" lrs = []\n",
" for p in optim.param_groups:\n",
" lrs.append(p['lr'])\n",
" if len(lrs) == 1:\n",
" return lrs[0]\n",
" else:\n",
" return lrs\n",
"\n",
"def minimize(T_init=None):\n",
" DeNoise = MyThresh(T_init) \n",
" optimizer = torch.optim.Adam(DeNoise.parameters(), lr=2e-0)\n",
" #scheduler = torch.optim.lr_scheduler.MultiStepLR(\n",
" # optimizer, milestones=[120, 180, 220], gamma=0.5)\n",
" criterion = torch.nn.MSELoss(reduction='elementwise_mean')\n",
" for step in range(250):\n",
" #scheduler.step()\n",
" optimizer.zero_grad()\n",
" Z = DeNoise(Y)\n",
" mse = torch.mean((X - Z)**2)\n",
" loss = criterion(Z,X)\n",
" if torch.isnan(mse):\n",
" raise ValueError('Nan encountered in training')\n",
" if step % 10 == 0:\n",
" print('Step [{:02d}]:\\tlr: {:.1e}\\tmse: {:.2f}'.format(\n",
" step, get_lr(optimizer), mse.item()))\n",
" loss.backward()\n",
" optimizer.step()\n",
" return DeNoise"
]
},
{
"cell_type": "code",
"execution_count": 29,
"metadata": {
"ExecuteTime": {
"end_time": "2018-12-03T16:55:04.130906Z",
"start_time": "2018-12-03T16:54:07.178500Z"
}
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Step [00]:\tlr: 2.0e+00\tmse: 382.41\n",
"Step [10]:\tlr: 2.0e+00\tmse: 117.28\n",
"Step [20]:\tlr: 2.0e+00\tmse: 88.03\n",
"Step [30]:\tlr: 2.0e+00\tmse: 84.91\n",
"Step [40]:\tlr: 2.0e+00\tmse: 83.98\n",
"Step [50]:\tlr: 2.0e+00\tmse: 83.60\n",
"Step [60]:\tlr: 2.0e+00\tmse: 83.27\n",
"Step [70]:\tlr: 2.0e+00\tmse: 83.11\n",
"Step [80]:\tlr: 2.0e+00\tmse: 83.02\n",
"Step [90]:\tlr: 2.0e+00\tmse: 82.98\n",
"Step [100]:\tlr: 2.0e+00\tmse: 82.95\n",
"Step [110]:\tlr: 2.0e+00\tmse: 82.93\n",
"Step [120]:\tlr: 2.0e+00\tmse: 82.92\n",
"Step [130]:\tlr: 2.0e+00\tmse: 82.90\n",
"Step [140]:\tlr: 2.0e+00\tmse: 82.89\n",
"Step [150]:\tlr: 2.0e+00\tmse: 82.88\n",
"Step [160]:\tlr: 2.0e+00\tmse: 82.87\n",
"Step [170]:\tlr: 2.0e+00\tmse: 82.86\n",
"Step [180]:\tlr: 2.0e+00\tmse: 82.86\n",
"Step [190]:\tlr: 2.0e+00\tmse: 82.85\n",
"Step [200]:\tlr: 2.0e+00\tmse: 82.85\n",
"Step [210]:\tlr: 2.0e+00\tmse: 82.84\n",
"Step [220]:\tlr: 2.0e+00\tmse: 82.84\n",
"Step [230]:\tlr: 2.0e+00\tmse: 82.83\n",
"Step [240]:\tlr: 2.0e+00\tmse: 82.83\n"
]
}
],
"source": [
"Denoise = minimize()"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Plot the new thresholds compared to the bayes threshold ones:"
]
},
{
"cell_type": "code",
"execution_count": 35,
"metadata": {
"ExecuteTime": {
"end_time": "2018-12-03T17:02:12.041885Z",
"start_time": "2018-12-03T17:02:11.733641Z"
},
"scrolled": false
},
"outputs": [
{
"data": {
"application/javascript": [
"/* Put everything inside the global mpl namespace */\n",
"window.mpl = {};\n",
"\n",
"\n",
"mpl.get_websocket_type = function() {\n",
" if (typeof(WebSocket) !== 'undefined') {\n",
" return WebSocket;\n",
" } else if (typeof(MozWebSocket) !== 'undefined') {\n",
" return MozWebSocket;\n",
" } else {\n",
" alert('Your browser does not have WebSocket support.' +\n",
" 'Please try Chrome, Safari or Firefox ≥ 6. ' +\n",
" 'Firefox 4 and 5 are also supported but you ' +\n",
" 'have to enable WebSockets in about:config.');\n",
" };\n",
"}\n",
"\n",
"mpl.figure = function(figure_id, websocket, ondownload, parent_element) {\n",
" this.id = figure_id;\n",
"\n",
" this.ws = websocket;\n",
"\n",
" this.supports_binary = (this.ws.binaryType != undefined);\n",
"\n",
" if (!this.supports_binary) {\n",
" var warnings = document.getElementById(\"mpl-warnings\");\n",
" if (warnings) {\n",
" warnings.style.display = 'block';\n",
" warnings.textContent = (\n",
" \"This browser does not support binary websocket messages. \" +\n",
" \"Performance may be slow.\");\n",
" }\n",
" }\n",
"\n",
" this.imageObj = new Image();\n",
"\n",
" this.context = undefined;\n",
" this.message = undefined;\n",
" this.canvas = undefined;\n",
" this.rubberband_canvas = undefined;\n",
" this.rubberband_context = undefined;\n",
" this.format_dropdown = undefined;\n",
"\n",
" this.image_mode = 'full';\n",
"\n",
" this.root = $('');\n",
" this._root_extra_style(this.root)\n",
" this.root.attr('style', 'display: inline-block');\n",
"\n",
" $(parent_element).append(this.root);\n",
"\n",
" this._init_header(this);\n",
" this._init_canvas(this);\n",
" this._init_toolbar(this);\n",
"\n",
" var fig = this;\n",
"\n",
" this.waiting = false;\n",
"\n",
" this.ws.onopen = function () {\n",
" fig.send_message(\"supports_binary\", {value: fig.supports_binary});\n",
" fig.send_message(\"send_image_mode\", {});\n",
" if (mpl.ratio != 1) {\n",
" fig.send_message(\"set_dpi_ratio\", {'dpi_ratio': mpl.ratio});\n",
" }\n",
" fig.send_message(\"refresh\", {});\n",
" }\n",
"\n",
" this.imageObj.onload = function() {\n",
" if (fig.image_mode == 'full') {\n",
" // Full images could contain transparency (where diff images\n",
" // almost always do), so we need to clear the canvas so that\n",
" // there is no ghosting.\n",
" fig.context.clearRect(0, 0, fig.canvas.width, fig.canvas.height);\n",
" }\n",
" fig.context.drawImage(fig.imageObj, 0, 0);\n",
" };\n",
"\n",
" this.imageObj.onunload = function() {\n",
" fig.ws.close();\n",
" }\n",
"\n",
" this.ws.onmessage = this._make_on_message_function(this);\n",
"\n",
" this.ondownload = ondownload;\n",
"}\n",
"\n",
"mpl.figure.prototype._init_header = function() {\n",
" var titlebar = $(\n",
" '');\n",
" var titletext = $(\n",
" '');\n",
" titlebar.append(titletext)\n",
" this.root.append(titlebar);\n",
" this.header = titletext[0];\n",
"}\n",
"\n",
"\n",
"\n",
"mpl.figure.prototype._canvas_extra_style = function(canvas_div) {\n",
"\n",
"}\n",
"\n",
"\n",
"mpl.figure.prototype._root_extra_style = function(canvas_div) {\n",
"\n",
"}\n",
"\n",
"mpl.figure.prototype._init_canvas = function() {\n",
" var fig = this;\n",
"\n",
" var canvas_div = $('');\n",
"\n",
" canvas_div.attr('style', 'position: relative; clear: both; outline: 0');\n",
"\n",
" function canvas_keyboard_event(event) {\n",
" return fig.key_event(event, event['data']);\n",
" }\n",
"\n",
" canvas_div.keydown('key_press', canvas_keyboard_event);\n",
" canvas_div.keyup('key_release', canvas_keyboard_event);\n",
" this.canvas_div = canvas_div\n",
" this._canvas_extra_style(canvas_div)\n",
" this.root.append(canvas_div);\n",
"\n",
" var canvas = $('');\n",
" canvas.addClass('mpl-canvas');\n",
" canvas.attr('style', \"left: 0; top: 0; z-index: 0; outline: 0\")\n",
"\n",
" this.canvas = canvas[0];\n",
" this.context = canvas[0].getContext(\"2d\");\n",
"\n",
" var backingStore = this.context.backingStorePixelRatio ||\n",
"\tthis.context.webkitBackingStorePixelRatio ||\n",
"\tthis.context.mozBackingStorePixelRatio ||\n",
"\tthis.context.msBackingStorePixelRatio ||\n",
"\tthis.context.oBackingStorePixelRatio ||\n",
"\tthis.context.backingStorePixelRatio || 1;\n",
"\n",
" mpl.ratio = (window.devicePixelRatio || 1) / backingStore;\n",
"\n",
" var rubberband = $('');\n",
" rubberband.attr('style', \"position: absolute; left: 0; top: 0; z-index: 1;\")\n",
"\n",
" var pass_mouse_events = true;\n",
"\n",
" canvas_div.resizable({\n",
" start: function(event, ui) {\n",
" pass_mouse_events = false;\n",
" },\n",
" resize: function(event, ui) {\n",
" fig.request_resize(ui.size.width, ui.size.height);\n",
" },\n",
" stop: function(event, ui) {\n",
" pass_mouse_events = true;\n",
" fig.request_resize(ui.size.width, ui.size.height);\n",
" },\n",
" });\n",
"\n",
" function mouse_event_fn(event) {\n",
" if (pass_mouse_events)\n",
" return fig.mouse_event(event, event['data']);\n",
" }\n",
"\n",
" rubberband.mousedown('button_press', mouse_event_fn);\n",
" rubberband.mouseup('button_release', mouse_event_fn);\n",
" // Throttle sequential mouse events to 1 every 20ms.\n",
" rubberband.mousemove('motion_notify', mouse_event_fn);\n",
"\n",
" rubberband.mouseenter('figure_enter', mouse_event_fn);\n",
" rubberband.mouseleave('figure_leave', mouse_event_fn);\n",
"\n",
" canvas_div.on(\"wheel\", function (event) {\n",
" event = event.originalEvent;\n",
" event['data'] = 'scroll'\n",
" if (event.deltaY < 0) {\n",
" event.step = 1;\n",
" } else {\n",
" event.step = -1;\n",
" }\n",
" mouse_event_fn(event);\n",
" });\n",
"\n",
" canvas_div.append(canvas);\n",
" canvas_div.append(rubberband);\n",
"\n",
" this.rubberband = rubberband;\n",
" this.rubberband_canvas = rubberband[0];\n",
" this.rubberband_context = rubberband[0].getContext(\"2d\");\n",
" this.rubberband_context.strokeStyle = \"#000000\";\n",
"\n",
" this._resize_canvas = function(width, height) {\n",
" // Keep the size of the canvas, canvas container, and rubber band\n",
" // canvas in synch.\n",
" canvas_div.css('width', width)\n",
" canvas_div.css('height', height)\n",
"\n",
" canvas.attr('width', width * mpl.ratio);\n",
" canvas.attr('height', height * mpl.ratio);\n",
" canvas.attr('style', 'width: ' + width + 'px; height: ' + height + 'px;');\n",
"\n",
" rubberband.attr('width', width);\n",
" rubberband.attr('height', height);\n",
" }\n",
"\n",
" // Set the figure to an initial 600x600px, this will subsequently be updated\n",
" // upon first draw.\n",
" this._resize_canvas(600, 600);\n",
"\n",
" // Disable right mouse context menu.\n",
" $(this.rubberband_canvas).bind(\"contextmenu\",function(e){\n",
" return false;\n",
" });\n",
"\n",
" function set_focus () {\n",
" canvas.focus();\n",
" canvas_div.focus();\n",
" }\n",
"\n",
" window.setTimeout(set_focus, 100);\n",
"}\n",
"\n",
"mpl.figure.prototype._init_toolbar = function() {\n",
" var fig = this;\n",
"\n",
" var nav_element = $('')\n",
" nav_element.attr('style', 'width: 100%');\n",
" this.root.append(nav_element);\n",
"\n",
" // Define a callback function for later on.\n",
" function toolbar_event(event) {\n",
" return fig.toolbar_button_onclick(event['data']);\n",
" }\n",
" function toolbar_mouse_event(event) {\n",
" return fig.toolbar_button_onmouseover(event['data']);\n",
" }\n",
"\n",
" for(var toolbar_ind in mpl.toolbar_items) {\n",
" var name = mpl.toolbar_items[toolbar_ind][0];\n",
" var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
" var image = mpl.toolbar_items[toolbar_ind][2];\n",
" var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
"\n",
" if (!name) {\n",
" // put a spacer in here.\n",
" continue;\n",
" }\n",
" var button = $('');\n",
" button.addClass('ui-button ui-widget ui-state-default ui-corner-all ' +\n",
" 'ui-button-icon-only');\n",
" button.attr('role', 'button');\n",
" button.attr('aria-disabled', 'false');\n",
" button.click(method_name, toolbar_event);\n",
" button.mouseover(tooltip, toolbar_mouse_event);\n",
"\n",
" var icon_img = $('');\n",
" icon_img.addClass('ui-button-icon-primary ui-icon');\n",
" icon_img.addClass(image);\n",
" icon_img.addClass('ui-corner-all');\n",
"\n",
" var tooltip_span = $('');\n",
" tooltip_span.addClass('ui-button-text');\n",
" tooltip_span.html(tooltip);\n",
"\n",
" button.append(icon_img);\n",
" button.append(tooltip_span);\n",
"\n",
" nav_element.append(button);\n",
" }\n",
"\n",
" var fmt_picker_span = $('');\n",
"\n",
" var fmt_picker = $('');\n",
" fmt_picker.addClass('mpl-toolbar-option ui-widget ui-widget-content');\n",
" fmt_picker_span.append(fmt_picker);\n",
" nav_element.append(fmt_picker_span);\n",
" this.format_dropdown = fmt_picker[0];\n",
"\n",
" for (var ind in mpl.extensions) {\n",
" var fmt = mpl.extensions[ind];\n",
" var option = $(\n",
" '', {selected: fmt === mpl.default_extension}).html(fmt);\n",
" fmt_picker.append(option)\n",
" }\n",
"\n",
" // Add hover states to the ui-buttons\n",
" $( \".ui-button\" ).hover(\n",
" function() { $(this).addClass(\"ui-state-hover\");},\n",
" function() { $(this).removeClass(\"ui-state-hover\");}\n",
" );\n",
"\n",
" var status_bar = $('');\n",
" nav_element.append(status_bar);\n",
" this.message = status_bar[0];\n",
"}\n",
"\n",
"mpl.figure.prototype.request_resize = function(x_pixels, y_pixels) {\n",
" // Request matplotlib to resize the figure. Matplotlib will then trigger a resize in the client,\n",
" // which will in turn request a refresh of the image.\n",
" this.send_message('resize', {'width': x_pixels, 'height': y_pixels});\n",
"}\n",
"\n",
"mpl.figure.prototype.send_message = function(type, properties) {\n",
" properties['type'] = type;\n",
" properties['figure_id'] = this.id;\n",
" this.ws.send(JSON.stringify(properties));\n",
"}\n",
"\n",
"mpl.figure.prototype.send_draw_message = function() {\n",
" if (!this.waiting) {\n",
" this.waiting = true;\n",
" this.ws.send(JSON.stringify({type: \"draw\", figure_id: this.id}));\n",
" }\n",
"}\n",
"\n",
"\n",
"mpl.figure.prototype.handle_save = function(fig, msg) {\n",
" var format_dropdown = fig.format_dropdown;\n",
" var format = format_dropdown.options[format_dropdown.selectedIndex].value;\n",
" fig.ondownload(fig, format);\n",
"}\n",
"\n",
"\n",
"mpl.figure.prototype.handle_resize = function(fig, msg) {\n",
" var size = msg['size'];\n",
" if (size[0] != fig.canvas.width || size[1] != fig.canvas.height) {\n",
" fig._resize_canvas(size[0], size[1]);\n",
" fig.send_message(\"refresh\", {});\n",
" };\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_rubberband = function(fig, msg) {\n",
" var x0 = msg['x0'] / mpl.ratio;\n",
" var y0 = (fig.canvas.height - msg['y0']) / mpl.ratio;\n",
" var x1 = msg['x1'] / mpl.ratio;\n",
" var y1 = (fig.canvas.height - msg['y1']) / mpl.ratio;\n",
" x0 = Math.floor(x0) + 0.5;\n",
" y0 = Math.floor(y0) + 0.5;\n",
" x1 = Math.floor(x1) + 0.5;\n",
" y1 = Math.floor(y1) + 0.5;\n",
" var min_x = Math.min(x0, x1);\n",
" var min_y = Math.min(y0, y1);\n",
" var width = Math.abs(x1 - x0);\n",
" var height = Math.abs(y1 - y0);\n",
"\n",
" fig.rubberband_context.clearRect(\n",
" 0, 0, fig.canvas.width, fig.canvas.height);\n",
"\n",
" fig.rubberband_context.strokeRect(min_x, min_y, width, height);\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_figure_label = function(fig, msg) {\n",
" // Updates the figure title.\n",
" fig.header.textContent = msg['label'];\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_cursor = function(fig, msg) {\n",
" var cursor = msg['cursor'];\n",
" switch(cursor)\n",
" {\n",
" case 0:\n",
" cursor = 'pointer';\n",
" break;\n",
" case 1:\n",
" cursor = 'default';\n",
" break;\n",
" case 2:\n",
" cursor = 'crosshair';\n",
" break;\n",
" case 3:\n",
" cursor = 'move';\n",
" break;\n",
" }\n",
" fig.rubberband_canvas.style.cursor = cursor;\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_message = function(fig, msg) {\n",
" fig.message.textContent = msg['message'];\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_draw = function(fig, msg) {\n",
" // Request the server to send over a new figure.\n",
" fig.send_draw_message();\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_image_mode = function(fig, msg) {\n",
" fig.image_mode = msg['mode'];\n",
"}\n",
"\n",
"mpl.figure.prototype.updated_canvas_event = function() {\n",
" // Called whenever the canvas gets updated.\n",
" this.send_message(\"ack\", {});\n",
"}\n",
"\n",
"// A function to construct a web socket function for onmessage handling.\n",
"// Called in the figure constructor.\n",
"mpl.figure.prototype._make_on_message_function = function(fig) {\n",
" return function socket_on_message(evt) {\n",
" if (evt.data instanceof Blob) {\n",
" /* FIXME: We get \"Resource interpreted as Image but\n",
" * transferred with MIME type text/plain:\" errors on\n",
" * Chrome. But how to set the MIME type? It doesn't seem\n",
" * to be part of the websocket stream */\n",
" evt.data.type = \"image/png\";\n",
"\n",
" /* Free the memory for the previous frames */\n",
" if (fig.imageObj.src) {\n",
" (window.URL || window.webkitURL).revokeObjectURL(\n",
" fig.imageObj.src);\n",
" }\n",
"\n",
" fig.imageObj.src = (window.URL || window.webkitURL).createObjectURL(\n",
" evt.data);\n",
" fig.updated_canvas_event();\n",
" fig.waiting = false;\n",
" return;\n",
" }\n",
" else if (typeof evt.data === 'string' && evt.data.slice(0, 21) == \"data:image/png;base64\") {\n",
" fig.imageObj.src = evt.data;\n",
" fig.updated_canvas_event();\n",
" fig.waiting = false;\n",
" return;\n",
" }\n",
"\n",
" var msg = JSON.parse(evt.data);\n",
" var msg_type = msg['type'];\n",
"\n",
" // Call the \"handle_{type}\" callback, which takes\n",
" // the figure and JSON message as its only arguments.\n",
" try {\n",
" var callback = fig[\"handle_\" + msg_type];\n",
" } catch (e) {\n",
" console.log(\"No handler for the '\" + msg_type + \"' message type: \", msg);\n",
" return;\n",
" }\n",
"\n",
" if (callback) {\n",
" try {\n",
" // console.log(\"Handling '\" + msg_type + \"' message: \", msg);\n",
" callback(fig, msg);\n",
" } catch (e) {\n",
" console.log(\"Exception inside the 'handler_\" + msg_type + \"' callback:\", e, e.stack, msg);\n",
" }\n",
" }\n",
" };\n",
"}\n",
"\n",
"// from http://stackoverflow.com/questions/1114465/getting-mouse-location-in-canvas\n",
"mpl.findpos = function(e) {\n",
" //this section is from http://www.quirksmode.org/js/events_properties.html\n",
" var targ;\n",
" if (!e)\n",
" e = window.event;\n",
" if (e.target)\n",
" targ = e.target;\n",
" else if (e.srcElement)\n",
" targ = e.srcElement;\n",
" if (targ.nodeType == 3) // defeat Safari bug\n",
" targ = targ.parentNode;\n",
"\n",
" // jQuery normalizes the pageX and pageY\n",
" // pageX,Y are the mouse positions relative to the document\n",
" // offset() returns the position of the element relative to the document\n",
" var x = e.pageX - $(targ).offset().left;\n",
" var y = e.pageY - $(targ).offset().top;\n",
"\n",
" return {\"x\": x, \"y\": y};\n",
"};\n",
"\n",
"/*\n",
" * return a copy of an object with only non-object keys\n",
" * we need this to avoid circular references\n",
" * http://stackoverflow.com/a/24161582/3208463\n",
" */\n",
"function simpleKeys (original) {\n",
" return Object.keys(original).reduce(function (obj, key) {\n",
" if (typeof original[key] !== 'object')\n",
" obj[key] = original[key]\n",
" return obj;\n",
" }, {});\n",
"}\n",
"\n",
"mpl.figure.prototype.mouse_event = function(event, name) {\n",
" var canvas_pos = mpl.findpos(event)\n",
"\n",
" if (name === 'button_press')\n",
" {\n",
" this.canvas.focus();\n",
" this.canvas_div.focus();\n",
" }\n",
"\n",
" var x = canvas_pos.x * mpl.ratio;\n",
" var y = canvas_pos.y * mpl.ratio;\n",
"\n",
" this.send_message(name, {x: x, y: y, button: event.button,\n",
" step: event.step,\n",
" guiEvent: simpleKeys(event)});\n",
"\n",
" /* This prevents the web browser from automatically changing to\n",
" * the text insertion cursor when the button is pressed. We want\n",
" * to control all of the cursor setting manually through the\n",
" * 'cursor' event from matplotlib */\n",
" event.preventDefault();\n",
" return false;\n",
"}\n",
"\n",
"mpl.figure.prototype._key_event_extra = function(event, name) {\n",
" // Handle any extra behaviour associated with a key event\n",
"}\n",
"\n",
"mpl.figure.prototype.key_event = function(event, name) {\n",
"\n",
" // Prevent repeat events\n",
" if (name == 'key_press')\n",
" {\n",
" if (event.which === this._key)\n",
" return;\n",
" else\n",
" this._key = event.which;\n",
" }\n",
" if (name == 'key_release')\n",
" this._key = null;\n",
"\n",
" var value = '';\n",
" if (event.ctrlKey && event.which != 17)\n",
" value += \"ctrl+\";\n",
" if (event.altKey && event.which != 18)\n",
" value += \"alt+\";\n",
" if (event.shiftKey && event.which != 16)\n",
" value += \"shift+\";\n",
"\n",
" value += 'k';\n",
" value += event.which.toString();\n",
"\n",
" this._key_event_extra(event, name);\n",
"\n",
" this.send_message(name, {key: value,\n",
" guiEvent: simpleKeys(event)});\n",
" return false;\n",
"}\n",
"\n",
"mpl.figure.prototype.toolbar_button_onclick = function(name) {\n",
" if (name == 'download') {\n",
" this.handle_save(this, null);\n",
" } else {\n",
" this.send_message(\"toolbar_button\", {name: name});\n",
" }\n",
"};\n",
"\n",
"mpl.figure.prototype.toolbar_button_onmouseover = function(tooltip) {\n",
" this.message.textContent = tooltip;\n",
"};\n",
"mpl.toolbar_items = [[\"Home\", \"Reset original view\", \"fa fa-home icon-home\", \"home\"], [\"Back\", \"Back to previous view\", \"fa fa-arrow-left icon-arrow-left\", \"back\"], [\"Forward\", \"Forward to next view\", \"fa fa-arrow-right icon-arrow-right\", \"forward\"], [\"\", \"\", \"\", \"\"], [\"Pan\", \"Pan axes with left mouse, zoom with right\", \"fa fa-arrows icon-move\", \"pan\"], [\"Zoom\", \"Zoom to rectangle\", \"fa fa-square-o icon-check-empty\", \"zoom\"], [\"\", \"\", \"\", \"\"], [\"Download\", \"Download plot\", \"fa fa-floppy-o icon-save\", \"download\"]];\n",
"\n",
"mpl.extensions = [\"eps\", \"jpeg\", \"pdf\", \"png\", \"ps\", \"raw\", \"svg\", \"tif\"];\n",
"\n",
"mpl.default_extension = \"png\";var comm_websocket_adapter = function(comm) {\n",
" // Create a \"websocket\"-like object which calls the given IPython comm\n",
" // object with the appropriate methods. Currently this is a non binary\n",
" // socket, so there is still some room for performance tuning.\n",
" var ws = {};\n",
"\n",
" ws.close = function() {\n",
" comm.close()\n",
" };\n",
" ws.send = function(m) {\n",
" //console.log('sending', m);\n",
" comm.send(m);\n",
" };\n",
" // Register the callback with on_msg.\n",
" comm.on_msg(function(msg) {\n",
" //console.log('receiving', msg['content']['data'], msg);\n",
" // Pass the mpl event to the overridden (by mpl) onmessage function.\n",
" ws.onmessage(msg['content']['data'])\n",
" });\n",
" return ws;\n",
"}\n",
"\n",
"mpl.mpl_figure_comm = function(comm, msg) {\n",
" // This is the function which gets called when the mpl process\n",
" // starts-up an IPython Comm through the \"matplotlib\" channel.\n",
"\n",
" var id = msg.content.data.id;\n",
" // Get hold of the div created by the display call when the Comm\n",
" // socket was opened in Python.\n",
" var element = $(\"#\" + id);\n",
" var ws_proxy = comm_websocket_adapter(comm)\n",
"\n",
" function ondownload(figure, format) {\n",
" window.open(figure.imageObj.src);\n",
" }\n",
"\n",
" var fig = new mpl.figure(id, ws_proxy,\n",
" ondownload,\n",
" element.get(0));\n",
"\n",
" // Call onopen now - mpl needs it, as it is assuming we've passed it a real\n",
" // web socket which is closed, not our websocket->open comm proxy.\n",
" ws_proxy.onopen();\n",
"\n",
" fig.parent_element = element.get(0);\n",
" fig.cell_info = mpl.find_output_cell(\"\");\n",
" if (!fig.cell_info) {\n",
" console.error(\"Failed to find cell for figure\", id, fig);\n",
" return;\n",
" }\n",
"\n",
" var output_index = fig.cell_info[2]\n",
" var cell = fig.cell_info[0];\n",
"\n",
"};\n",
"\n",
"mpl.figure.prototype.handle_close = function(fig, msg) {\n",
" var width = fig.canvas.width/mpl.ratio\n",
" fig.root.unbind('remove')\n",
"\n",
" // Update the output cell to use the data from the current canvas.\n",
" fig.push_to_output();\n",
" var dataURL = fig.canvas.toDataURL();\n",
" // Re-enable the keyboard manager in IPython - without this line, in FF,\n",
" // the notebook keyboard shortcuts fail.\n",
" IPython.keyboard_manager.enable()\n",
" $(fig.parent_element).html('');\n",
" fig.close_ws(fig, msg);\n",
"}\n",
"\n",
"mpl.figure.prototype.close_ws = function(fig, msg){\n",
" fig.send_message('closing', msg);\n",
" // fig.ws.close()\n",
"}\n",
"\n",
"mpl.figure.prototype.push_to_output = function(remove_interactive) {\n",
" // Turn the data on the canvas into data in the output cell.\n",
" var width = this.canvas.width/mpl.ratio\n",
" var dataURL = this.canvas.toDataURL();\n",
" this.cell_info[1]['text/html'] = '';\n",
"}\n",
"\n",
"mpl.figure.prototype.updated_canvas_event = function() {\n",
" // Tell IPython that the notebook contents must change.\n",
" IPython.notebook.set_dirty(true);\n",
" this.send_message(\"ack\", {});\n",
" var fig = this;\n",
" // Wait a second, then push the new image to the DOM so\n",
" // that it is saved nicely (might be nice to debounce this).\n",
" setTimeout(function () { fig.push_to_output() }, 1000);\n",
"}\n",
"\n",
"mpl.figure.prototype._init_toolbar = function() {\n",
" var fig = this;\n",
"\n",
" var nav_element = $('')\n",
" nav_element.attr('style', 'width: 100%');\n",
" this.root.append(nav_element);\n",
"\n",
" // Define a callback function for later on.\n",
" function toolbar_event(event) {\n",
" return fig.toolbar_button_onclick(event['data']);\n",
" }\n",
" function toolbar_mouse_event(event) {\n",
" return fig.toolbar_button_onmouseover(event['data']);\n",
" }\n",
"\n",
" for(var toolbar_ind in mpl.toolbar_items){\n",
" var name = mpl.toolbar_items[toolbar_ind][0];\n",
" var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
" var image = mpl.toolbar_items[toolbar_ind][2];\n",
" var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
"\n",
" if (!name) { continue; };\n",
"\n",
" var button = $('');\n",
" button.click(method_name, toolbar_event);\n",
" button.mouseover(tooltip, toolbar_mouse_event);\n",
" nav_element.append(button);\n",
" }\n",
"\n",
" // Add the status bar.\n",
" var status_bar = $('');\n",
" nav_element.append(status_bar);\n",
" this.message = status_bar[0];\n",
"\n",
" // Add the close button to the window.\n",
" var buttongrp = $('');\n",
" var button = $('');\n",
" button.click(function (evt) { fig.handle_close(fig, {}); } );\n",
" button.mouseover('Stop Interaction', toolbar_mouse_event);\n",
" buttongrp.append(button);\n",
" var titlebar = this.root.find($('.ui-dialog-titlebar'));\n",
" titlebar.prepend(buttongrp);\n",
"}\n",
"\n",
"mpl.figure.prototype._root_extra_style = function(el){\n",
" var fig = this\n",
" el.on(\"remove\", function(){\n",
"\tfig.close_ws(fig, {});\n",
" });\n",
"}\n",
"\n",
"mpl.figure.prototype._canvas_extra_style = function(el){\n",
" // this is important to make the div 'focusable\n",
" el.attr('tabindex', 0)\n",
" // reach out to IPython and tell the keyboard manager to turn it's self\n",
" // off when our div gets focus\n",
"\n",
" // location in version 3\n",
" if (IPython.notebook.keyboard_manager) {\n",
" IPython.notebook.keyboard_manager.register_events(el);\n",
" }\n",
" else {\n",
" // location in version 2\n",
" IPython.keyboard_manager.register_events(el);\n",
" }\n",
"\n",
"}\n",
"\n",
"mpl.figure.prototype._key_event_extra = function(event, name) {\n",
" var manager = IPython.notebook.keyboard_manager;\n",
" if (!manager)\n",
" manager = IPython.keyboard_manager;\n",
"\n",
" // Check for shift+enter\n",
" if (event.shiftKey && event.which == 13) {\n",
" this.canvas_div.blur();\n",
" event.shiftKey = false;\n",
" // Send a \"J\" for go to next cell\n",
" event.which = 74;\n",
" event.keyCode = 74;\n",
" manager.command_mode();\n",
" manager.handle_keydown(event);\n",
" }\n",
"}\n",
"\n",
"mpl.figure.prototype.handle_save = function(fig, msg) {\n",
" fig.ondownload(fig, null);\n",
"}\n",
"\n",
"\n",
"mpl.find_output_cell = function(html_output) {\n",
" // Return the cell and output element which can be found *uniquely* in the notebook.\n",
" // Note - this is a bit hacky, but it is done because the \"notebook_saving.Notebook\"\n",
" // IPython event is triggered only after the cells have been serialised, which for\n",
" // our purposes (turning an active figure into a static one), is too late.\n",
" var cells = IPython.notebook.get_cells();\n",
" var ncells = cells.length;\n",
" for (var i=0; i= 3 moved mimebundle to data attribute of output\n",
" data = data.data;\n",
" }\n",
" if (data['text/html'] == html_output) {\n",
" return [cell, data, j];\n",
" }\n",
" }\n",
" }\n",
" }\n",
"}\n",
"\n",
"// Register the function which deals with the matplotlib target/channel.\n",
"// The kernel may be null if the page has been refreshed.\n",
"if (IPython.notebook.kernel != null) {\n",
" IPython.notebook.kernel.comm_manager.register_target('matplotlib', mpl.mpl_figure_comm);\n",
"}\n"
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/html": [
"![](\"data:image/png;base64,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\")
"
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"colors = 'RGB'\n",
"c = colors.find('R')\n",
"T = bayes_thresh(im_noise)\n",
"\n",
"plt.figure(figsize=(10,10))\n",
"gs = gridspec.GridSpec(3,3)\n",
"gs.update(left=0.1, right=0.95, wspace=0.05, hspace=.2, top=0.95)\n",
"for j in range(J):\n",
" for b, band in enumerate(['LH', 'HL', 'HH']): \n",
" if b == 0: \n",
" ax = plt.subplot(gs[j,b])\n",
" ax.yaxis.set_major_formatter(FormatStrFormatter('%.2f'))\n",
" ax0 = ax \n",
" else:\n",
" ax = plt.subplot(gs[j,b])\n",
" plt.setp(ax.get_yticklabels(), visible=False)\n",
" \n",
" ax.plot(Ts[j,b,c], mses[j,b,c])\n",
" if j == 0 and b == 1: \n",
" ax.axvline(x=T[j,b,c], ls='--',c='r',\n",
" label='Bayes calculated thresh') \n",
" ax.axvline(x=Denoise.shrinkers[2].t[c,1], ls='--', c='g', \n",
" label='Backprop calculated thresh')\n",
" ax.legend(framealpha=1)\n",
" else:\n",
" ax.axvline(x=T[j,b,c], ls='--',c='r')\n",
" ax.axvline(x=Denoise.shrinkers[J-1-j].t[c,b], ls='--', c='g')\n",
" \n",
" ax.set_title('Scale {} {}'.format(J-j, band))\n",
" ax.set_xlim(0, max(Ts[j,b,c,-1], T[j,b,c]*1.1))\n",
" "
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"It works! We have learned the optimal thresholds via backpropagation. As mentioned at the beginning of this notebook however, this is still not ideal as we needed the noiseless image to calculate the MSE."
]
}
],
"metadata": {
"kernelspec": {
"display_name": "Python",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.6.5"
},
"toc": {
"colors": {
"hover_highlight": "#DAA520",
"navigate_num": "#000000",
"navigate_text": "#333333",
"running_highlight": "#FF0000",
"selected_highlight": "#FFD700",
"sidebar_border": "#EEEEEE",
"wrapper_background": "#FFFFFF"
},
"moveMenuLeft": true,
"nav_menu": {
"height": "45px",
"width": "252px"
},
"navigate_menu": true,
"number_sections": true,
"sideBar": true,
"threshold": 4,
"toc_cell": true,
"toc_position": {
"height": "974px",
"left": "0px",
"right": "1552.25px",
"top": "133px",
"width": "367.75px"
},
"toc_section_display": "block",
"toc_window_display": false,
"widenNotebook": false
}
},
"nbformat": 4,
"nbformat_minor": 2
}