K-means Clustering for Image Segmentation to find sum of cluster's pixels

Hello MATLAB community,

I've been using K-mean clustering for image segmentation; this is electron beam data. I've used to code below:

img = data.img(100:500,400:900); % Select the relevant part of the image
[L,Centers] = imsegkmeans(img,8); % Perform image segmentation with 8 centroids
 B = labeloverlay(img,L);
imshow(B); % Display it

This code produces the following output:

There are 8 "islands" that represent a quantity that needs to be calculated. What I need to do is to calculated the pixels that account for these "islands". Any suggestions how I may be able to do this?

Answer 1

[Pratyush Roy]

Hi Sebastian,

Assuming that the goal is finding the pixels corresponding to a particular cluster, one can use the "find" function to get the indices of the pixels corresponding to a particular cluster.

Here's a code snippet that can be used to get the pixel indices

[R1,C1] = find(L,1);% R1 gives us the row indices and C1 gives us the column indices for the points with label 1.
[R2,C2] = find(L,2);% Similarly we can get for pixels for label 2

In this way we can get the indices for the pixels for all the 8 clusters. To get the pixel intensity values,one can append all the intensity values to an empty array "arr" to get the intensity values of the pixels belonging to the same cluster.

arr = [];% Stores pixel intensity values for cluster 1. One can have a different empty array for a different cluster. 
for i=1:length(R1)
    J = I(R1(i),C1(i));
    arr = [arr J];
end

You can go through the following documentation link for further help:

https://www.mathworks.com/help/matlab/ref/find.html

[Sebastian Bustillo]

Thank you so much for your help! I was wondering if I take the sum of the pixels from the original img matrix or from B (the output of overlay). Sorry if this is a dumb question but I'm very new to image segmenation with MATLAB.