# GraphImage2List for multi-colored graph-image-data **URL:** \ **Category:** Staff Picks\ **Tags:** image-processing, graphics-and-visualization, interface-construction, machine-learning, staff-pick\ **Created:** [May 22, 2017, 8:20pm UTC](https://community.wolfram.com/t/graphimage2list-for-multi-colored-graph-image-data/10572 "2017-05-22T20:20:01Z")\ **Posts on this page:** 2\ **Page:** 1
**Author:** ![okazakikotaro](https://yyz1.discourse-cdn.com/wolfram/user_avatar/community.wolfram.com/okazakikotaro/32/14323_2.png) [@okazakikotaro](https://community.wolfram.com/u/okazakikotaro)\ **Post date:** [May 22, 2017, 8:20pm UTC](https://community.wolfram.com/t/graphimage2list-for-multi-colored-graph-image-data/10572/1 "2017-05-22T20:20:01Z")
In my previous post [GraphImage2List](https://wolfram.discourse.group/t/10462/1) , I created the tool that changes graph-image-data to numeric-data in order to do a Machine Learning using graph-image-data. However, it can’t deal with multi-colored graph-image-data. So I create a new tool by **ImageGraphics** function. ImageGraphics is one of my favorite new functions in [Wolfram Language](http://reference.wolfram.com/language/ref/ImageGraphics.html) 11.1. ## **Goal** The left is the original graph(image data) like CPU Utilization of five virtual machines. The right is the graph(ListPlot) of yellow-green line picked up by the new tool. ![screenshot.0.jpg](https://global.discourse-cdn.com/wolfram/original/3X/1/f/1f06fea9f6ab93084d39588565e8fdd1e5514b49.jpeg) The tool comes in three steps. ## **Step 1** In the first step, My MakeMasking function masks unnecessary areas around the original image data like plot label, frame and ticks. It outputs “maskedimage”. ``` MakeMasking[img_] := Module[{}, size = {sizex, sizey} = ImageDimensions[img]; backimg = Image[Table[1, {sizey}, {sizex}]]; Manipulate[ Grid[{{"input image", "masked image"}, {Show[img, ImageSize -> size], Show[ maskedimg = ImageCompose[backimg, ImageTrim[ img, {{left, bottom}, {right, top}}], {left + right, bottom + top}/2], ImageSize -> size]}}], Row[{Control[{left, 0, sizex, 1}], Control[{{right, sizex}, 0, sizex, 1}]}, " "], Row[{Control[{bottom, 0, sizey, 1}], Control[{{top, sizey}, 0, sizey, 1}]}, " "] ] ] ``` The left is the original. The right is masked image. ![9670screenshot.2.jpg](https://global.discourse-cdn.com/wolfram/original/2X/4/46507c36f1ec5cebd6f6d2261c0445d10e48bea4.jpeg) ## **Step 2** In this step, My SelectColors function selects the areas of two near colors in ImageGraphics output of “maskedimage”. ImageGraphics function returns the content of image. The colors are below. ``` maskedimggraphics = ImageGraphics[maskedimg, PlotRangePadding -> None]; maskedimggraphics[[1, 2, #, 1]] & /@ Range[Length[maskedimggraphics[[1, 2]]]] ``` ![1885screenshot.8.jpg](https://global.discourse-cdn.com/wolfram/optimized/2X/8/8a0ae2f0d5fad858df3d858b6388d02a261078a0_2_932x25.jpeg) My SelectColors function uses the content and outputs the selected area as “selectpts”. ``` SelectColors[img_, maskedimggraphics_] := Module[{}, {sizex, sizey} = size = ImageDimensions[img] // N; frame = {{0., 0.}, {0., sizey}, {sizex, sizey}, {sizex, 0.}}; l = Length[maskedimggraphics[[1, 1]]]; colors = maskedimggraphics[[1, 2, #, 1]] & /@ Range[Length[maskedimggraphics[[1, 2]]]]; Manipulate[ Grid[{{"input image", "selected image"}, {Show[img, ImageSize -> ImageDimensions[img]], Graphics[ GraphicsComplex[Join[maskedimggraphics[[1, 1]], frame], Join[{LABColor[ 0.9988949153893414, -3.6790844387894895`*^-6, 0.00042430735605277474`], FilledCurve[{{Line[{l + 1, l + 2, l + 3, l + 4}]}}]}, selectpts = FirstCase[maskedimggraphics[[1, 2]], {#, ___}, Nothing] & /@ {color1, color2}]], ImageSize -> size]}}], {{color1, colors[[2]]}, colors, ControlType -> RadioButton, Method -> "Queued"}, {{color2, colors[[2]]}, colors, ControlType -> RadioButton, Method -> "Queued"}, SynchronousUpdating -> False] ] ``` The left is the original. The right is selected image. ![10670screenshot.3.jpg](https://global.discourse-cdn.com/wolfram/original/3X/4/7/4709a8da5c3b62c1c6aad65acb29ad5b822e02de.jpeg) ## **Step 3** You can make list in the next process by using my GetList2 function. 1. Add/del points with alt+click(WINDOWS)/cmd+click(MAC) if necessary 2. Set x and y values(Min, Max, Accuracy) of red points 3. Click Calculate button ![1651screenshot.4.jpg](https://global.discourse-cdn.com/wolfram/original/2X/e/e0a7437a43e09431b544f0601fc01a41d98af9c1.jpeg) Set x and y values(Min, Max, Accuracy) of red points and click. ![9510screenshot.5.jpg](https://global.discourse-cdn.com/wolfram/original/3X/f/7/f71bdd5df65052f671430c4326ea55d086b6b3ef.jpeg) Then it outputs “list”. ``` list ``` ![6445screenshot.12.jpg](https://global.discourse-cdn.com/wolfram/optimized/2X/7/7ba15a1c94a05c98fafd202b6ad4b3e1a1afe8ab_2_932x48.jpeg) ListPlot of the list is below. ``` ListPlot[Style[list, RGBColor[204/255, 204/255, 0]], Joined -> True] ``` ![2322screenshot.7.jpg](https://global.discourse-cdn.com/wolfram/original/3X/f/c/fc6768129287adc8339493f06ce0904e6b7249f5.jpeg) ## **Differences** When GetList2 converts area selected in Step 2. to coordinates, there are some points of the same x coordinate. So my locator2coordinate2 function outputs 4 lists of y coordinate, high, low, middle and combination as nearyH, nearyL, nearyM and nearyC. As I show below, nearyC seems to be better than others. I create this image data below. ``` data = {{6, 5, 33, 36, 9, 11, 23, 29, 34, 26, 3, 6, 26, 35, 21, 6, 26, 33, 20, 16, 30, 6, 1, 6}, {41, 34, 43, 60, 33, 38, 54, 43, 29, 59, 45, 34, 42, 55, 42, 26, 59, 20, 20, 41, 41, 47, 28, 52}, {58, 55, 61, 56, 40, 47, 50, 72, 72, 66, 69, 69, 78, 75, 70, 66, 56, 76, 66, 43, 47, 79, 56, 49}, {88, 96, 84, 62, 69, 67, 61, 60, 94, 76, 75, 70, 69, 86, 68, 61, 72, 91, 89, 71, 69, 83, 88, 75}, {17, 9, 23, 19, 23, 47, 45, 30, 82, 88, 58, 24, 59, 61, 17, 82, 95, 83, 40, 81, 68, 5, 40, 7}}; graph = DateListPlot[data, {2017, 5, 20, 0}, Frame -> True, FrameLabel -> {"Time", "CPU Utilization %"}, PlotStyle -> 96]; img = Rasterize[graph] ``` Mean and Variance of the difference between the true list(data) and calculated 4 list(nearyH, nearyL, nearyM and nearyC) are below and nearyC is the best of all. ``` Grid[ Join[{{"", "nearyH", "nearyM", "nearyL", "nearyc"}}, Join[{{"Mean", "Variance"}}, {Mean[#], Variance[#]} & /@ {nearyH - #, nearyM - #, nearyL - #, nearyC - #} &@data[[5]] // N] // Transpose] // Transpose, Frame -> All] ``` ![1844screenshot.9.jpg](https://global.discourse-cdn.com/wolfram/original/2X/4/460221f5d5ab6c0f4cf5c01f1181fc63f433e380.jpeg) ``` ListPlot[{Legended[Style[data[[5]], RGBColor[204/255, 204/255, 0]], "data[[5]]"], Legended[Style[nearyC, Red], "nearyC"]}, Joined -> {True, False}, PlotRange -> {0, 100}] ``` ![6267screenshot.6.jpg](https://global.discourse-cdn.com/wolfram/original/1X/b82e3ba78f3b1fed156059044d73818c07aaf9fe.jpeg) ---
**Author:** ![wolframcommunity](https://yyz1.discourse-cdn.com/wolfram/user_avatar/community.wolfram.com/wolframcommunity/32/8335_2.png) [@wolframcommunity](https://community.wolfram.com/u/wolframcommunity)\ **Post date:** [September 22, 2017, 2:06pm UTC](https://community.wolfram.com/t/graphimage2list-for-multi-colored-graph-image-data/10572/2 "2017-09-22T14:06:54Z")
![2015-11-02_10-07-11.png](https://global.discourse-cdn.com/wolfram/original/3X/5/0/50e59c9ead62f729e7130d3ea1482555a75602b6.png) - Congratulations! This post is now a [Staff Pick](http://community.wolfram.com/content?curTag=staff+picks) as distinguished by a badge on [your profile](http://community.wolfram.com/web/okazakikotaro)! Thank you, keep it coming!