Displaying kmean result with specific colors to specific clusters

Question

I applied k-mean clustering on a preprocessed image using the following matlab code

%B - input image
C=rgb2gray(B);
[idx centroids]=kmeans(double(C(:)),4);
imseg = zeros(size(C,1),size(C,2));
for i=1:max(idx)
     imseg(idx==i)=i;
end
i=mat2gray(imseg);
% i - output image

Every time I display the output, color assigned to the output images changes. How can I give a specific color to cluster1, cluster2, cluster3 and cluster4.

you can do it in the line you display the output, unfortunately you havent added it to the code so we cant help... — Ander Biguri

Shai Shai · Accepted Answer · 2015-04-27T08:20:59

You can use a colormap. Let R1, B1 and G1 be the RGB values you want to dispaly the first cluster in (values in range [0..1]), and R2 be the red channel value for the second cluster and so on... Then your color map is:

cmp = [R1 G1 B1;
       ...
       R4 G4 B4];

Now,

[idx centroids] = kmeans(double(C(:)),4);
imseg = reshape( idx, size(C) ); %// reshape
figure; imagesc( imseg );colormap( cmp ); %// that's it!

PS,
It is best not to use i as a variable name in Matlab.

Displaying kmean result with specific colors to specific clusters

1 Answers