Using marching squares algorithm to trace the contour of an image
Did you ever wonder how graphic softwares can trace the contour of an image? No? Actually, I never cared about it, but since I needed to do it for a more complex project, I discovered the magic world of marching squares algorithm to trace the contour of an image.
The idea of the process is simple:
1) Find a pixel placed on the edge of the image (this has nothing to do with marching squares, which assumes you already found such pixel). This will be the current pixel to analyze
2) Consider a 2×2 pixel square which includes the current pixel which generally is located in the top left or the right bottom of the square.
3) At this time you have 4 pixels, and each one of them can be transparent or opaque. So we will have 16 possible 2×2 squares, although the cases with all transparent pixels or all opaque pixels will never occur as we are moving around the edge of the image.
4) According to the number and position of opaque pixels in the 2×2 square, we can guess the direction of the contour, move the current pixel in such direction, and continue from step 2 until you reach again the pixel found at step 1
Now, how can we find the starting pixel needed at step 1? This is just a matter of brute force, we need to scan the image pixel by pixel until we find an opaque pixel.
Let’s see this fully commented code:
|
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 |
package { import flash.display.Sprite; import flash.display.BitmapData; import flash.display.Bitmap; import flash.geom.Matrix; import flash.geom.Point; public class Main extends Sprite { private var bitmapData:BitmapData=new BitmapData(640,480,true,0x00000000); // tolerance is the amount of alpha for a pixel to be considered solid private var tolerance:Number=0x01; public function Main() { // adding a png image with transparency bitmapData.draw(new Logo(278,429),new Matrix(1,0,0,1,100,40)); var bitmap:Bitmap=new Bitmap(bitmapData); addChild(bitmap); // at the end of this function, marchingVector will contain the points tracing the contour var marchingVector:Vector.<Point>=marchingSquares(bitmapData); } public function marchingSquares(bitmapData:BitmapData):Vector.<Point> { var contourVector:Vector.<Point> = new Vector.<Point>(); // this is the canvas we'll use to draw the contour var canvas:Sprite=new Sprite(); addChild(canvas); canvas.graphics.lineStyle(2,0x00ff00); // getting the starting pixel var startPoint:Point=getStartingPixel(bitmapData); // if we found a starting pixel we can begin if (startPoint!=null) { // moving the graphic pen to the starting pixel canvas.graphics.moveTo(startPoint.x,startPoint.y); // pX and pY are the coordinates of the starting point var pX:Number=startPoint.x; var pY:Number=startPoint.y; // stepX and stepY can be -1, 0 or 1 and represent the step in pixels to reach // next contour point var stepX:Number; var stepY:Number; // we also need to save the previous step, that's why we use prevX and prevY var prevX:Number; var prevY:Number; // closedLoop will be true once we traced the full contour var closedLoop:Boolean=false; while (!closedLoop) { // the core of the script is getting the 2x2 square value of each pixel var squareValue:Number=getSquareValue(pX,pY); switch (squareValue) { /* going UP with these cases: +---+---+ +---+---+ +---+---+ | 1 | | | 1 | | | 1 | | +---+---+ +---+---+ +---+---+ | | | | 4 | | | 4 | 8 | +---+---+ +---+---+ +---+---+ */ case 1 : case 5 : case 13 : stepX=0; stepY=-1; break; /* going DOWN with these cases: +---+---+ +---+---+ +---+---+ | | | | | 2 | | 1 | 2 | +---+---+ +---+---+ +---+---+ | | 8 | | | 8 | | | 8 | +---+---+ +---+---+ +---+---+ */ case 8 : case 10 : case 11 : stepX=0; stepY=1; break; /* going LEFT with these cases: +---+---+ +---+---+ +---+---+ | | | | | | | | 2 | +---+---+ +---+---+ +---+---+ | 4 | | | 4 | 8 | | 4 | 8 | +---+---+ +---+---+ +---+---+ */ case 4 : case 12 : case 14 : stepX=-1; stepY=0; break; /* going RIGHT with these cases: +---+---+ +---+---+ +---+---+ | | 2 | | 1 | 2 | | 1 | 2 | +---+---+ +---+---+ +---+---+ | | | | | | | 4 | | +---+---+ +---+---+ +---+---+ */ case 2 : case 3 : case 7 : stepX=1; stepY=0; break; case 6 : /* special saddle point case 1: +---+---+ | | 2 | +---+---+ | 4 | | +---+---+ going LEFT if coming from UP else going RIGHT */ if (prevX==0&&prevY==-1) { stepX=-1; stepY=0; } else { stepX=1; stepY=0; } break; case 9 : /* special saddle point case 2: +---+---+ | 1 | | +---+---+ | | 8 | +---+---+ going UP if coming from RIGHT else going DOWN */ if (prevX==1&&prevY==0) { stepX=0; stepY=-1; } else { stepX=0; stepY=1; } break; } // moving onto next point pX+=stepX; pY+=stepY; // saving contour point contourVector.push(new Point(pX, pY)); prevX=stepX; prevY=stepY; // drawing the line canvas.graphics.lineTo(pX,pY); // if we returned to the first point visited, the loop has finished if (pX==startPoint.x&&pY==startPoint.y) { closedLoop=true; } } } return contourVector; } private function getStartingPixel(bitmapData:BitmapData):Point { // finding the starting pixel is a matter of brute force, we need to scan // the image pixel by pixel until we find a non-transparent pixel var zeroPoint:Point=new Point(0,0); var offsetPoint:Point=new Point(0,0); for (var i:Number=0; i<bitmapData.height; i++) { for (var j:Number=0; j<bitmapData.width; j++) { offsetPoint.x=j; offsetPoint.y=i; if (bitmapData.hitTest(zeroPoint,tolerance,offsetPoint)) { return offsetPoint; } } } return null; } private function getSquareValue(pX:Number,pY:Number):Number { /* checking the 2x2 pixel grid, assigning these values to each pixel, if not transparent +---+---+ | 1 | 2 | +---+---+ | 4 | 8 | <- current pixel (pX,pY) +---+---+ */ var squareValue:Number=0; // checking upper left pixel if (getAlphaValue(bitmapData.getPixel32(pX-1,pY-1))>=tolerance) { squareValue+=1; } // checking upper pixel if (getAlphaValue(bitmapData.getPixel32(pX,pY-1))>tolerance) { squareValue+=2; } // checking left pixel if (getAlphaValue(bitmapData.getPixel32(pX-1,pY))>tolerance) { squareValue+=4; } // checking the pixel itself if (getAlphaValue(bitmapData.getPixel32(pX,pY))>tolerance) { squareValue+=8; } return squareValue; } private function getAlphaValue(n:Number):Number { // given an ARGB color value, returns the alpha 0 -> 255 return n >> 24 & 0xFF; } } } |
And this is the result:
The green line is the contour traced by the algorithm. Next time, I am showing you how to use this algorithm with Box2D to create interesting things.
(12 votes, average: 4.75 out of 5)
Loading ...
They can be easily customized to meet the unique requirements of your project.
This post has 7 comments
ladeng6666
Amazing, we can use the vector got from marchingSquares method together with b2Separator to create a complicated polygon, and what’s more, we could load a customer Image on fly. am I right?
Carlos Ribeiro
My God, you are amazing.
Emanuele Feronato
@ladeng6666: yes, using b2Separator (http://www.emanueleferonato.com/2011/09/12/create-non-convex-complex-shapes-with-box2d/) or triangulation (http://www.emanueleferonato.com/2012/06/27/polygon-triangulation-decomposition-of-a-polygon-into-triangles-with-as3/)
jamal
amazing tutorial as usual , thank u
narongrit
what direction to move
when
square value = 0 and square value = 15
Emanuele Feronato
@narongrit: if the scripts executes correctly, you’ll never find 0 or 15 values
thx
amazing,thank u!