watershed algorithm in matlab

Glove picture Glove · Jul 14, 2011 · Viewed 10k times · Source

anyone knows how to write a function in matlab to segment the cells and compute the average cell area using the watershed algorithm? any help would be much appreciated. Thank you!

Here is an image of yeast cells

yeast cells

Answer

Jonas picture Jonas · Jul 14, 2011

Here's one way to segment the image using watershed. There's plenty more you could do (e.g. fuse cells with two nuclei if they haven't completed cytokinesis yet), but the steps below should give you a first idea.

(1) Determine cell-background threshold, cell-nucleus threshold

%# read image
img = imread('http://i.stack.imgur.com/nFDkX.png');
%# normalize to 0...1
imgN = double(img-min(img(:)))/(max(img(:)-min(img(:))));
th1=graythresh(imgN);
th2 = graythresh(imgN(imgN>th1));

cellMsk = imgN>th1;
nucMsk = imgN>th2;

figure,imshow(cellMsk+nucMsk,[])

enter image description here

(2) Smooth the raw image (to avoid oversegmentation) and impose nuclei as minima

[xx,yy]=ndgrid(-5:5,-5:5);
gf = exp((-xx.^2-yy.^2)/20);
filtImg = conv2(imgN,gf,'same');

figure,imshow(filtImg,[])

filtImgM = imimposemin(-filtImg,nucMsk);

enter image description here

(3) Watershed, mask cells, and display

ws = watershed(filtImgM);
ws(~cellMsk) = 0;

lblImg = bwlabel(ws);

figure,imshow(label2rgb(lblImg,'jet','k','shuffle'));

enter image description here

(4) Now you can use REGIONPROPS on the labeled image to extract the statistics you want.