Finding locations of images clusters centroids using imsegkmeans

I use imsegkmeans to cluster images and I need to know the centroid of each cluster. So I write this line:

[Labels,centers] = imsegkmeans(div_img,4);

My image is an RGB 250-250 pixels and I am clustering it into 4 clusters, so I got a 4-3 matrix which matches the explaination in the documentation.

centers

centers =

  4×3 uint8 matrix

   25   25   29
   15    9   13
   64   58   54
   34   13   14

My understading is that the first row [25,25,29] contains the centroid for the first cluster and it has 3 values for R, G and B channels. However, I don't understand how this is a location? I need indices for row and column for the centroid of each cluster

Answer 1

[Shubham Rawat]

Hi Hassan,

In the imsegkmeans function you will have centroid values of the colors not the positions.

I have created a custom code which will generate Centroid positions of each cluster with the help of Label matrix:

% input Image(RGB) and number of clusters k
% output clusteIndices k*2 size
function clusterIndices = clusterIndices(I,k)
    % L is the Label matrix here
    [L,~] = imsegkmeans(I,k);
    
    % initialization 
    clusterIndices = zeros(k,2);
    sumx = zeros(k,2); 
    sumy = zeros(k,2);
    num = zeros(k,1);
    
    for c = 1:k
        for i = 1:size(L,1)
            for j = 1:size(L,2)
                if L(i,j) == c
                    sumx(c) = sumx(c)+i;
                    sumy(c) = sumy(c)+j;
                    num(c) = num(c)+1;
                end
            end
        end
        clusterIndices(c,1) = sumx(c)/num(c); 
        clusterIndices(c,2) = sumy(c)/num(c);
    end
end

Hope this Helps!