How can I separate my a green channel of the image using k-means clustering?

I have been working on a project that separates leaf color(green) from the remaining background and plots it. The only problem is that whenever I do it each cluster comes out random (one cluster would equal another and vice versa) here is my code

    he = imread('6.jpg');
%imshow(he), title('LEAF');
text(size(he,2),size(he,1)+15,...
     'Image courtesy of *******, ', ...
     'FontSize',7,'HorizontalAlignment','right');
 cform = makecform('srgb2lab');
lab_he = applycform(he,cform);
ab = double(lab_he(:,:,2:3));
nrows = size(ab,1);
ncols = size(ab,2);
ab = reshape(ab,nrows*ncols,2);
nColors = 3;
% repeat the clustering 3 times to avoid local minima
[cluster_idx, cluster_center] = kmeans(ab,nColors,'distance','sqEuclidean', ...
                                      'Replicates',3);
pixel_labels = reshape(cluster_idx,nrows,ncols);
%imshow(pixel_labels,[]), title('image labeled by cluster index');
segmented_images = cell(1,3);
rgb_label = repmat(pixel_labels,[1 1 3]);
for k = 1:nColors
    color = he;
    color(rgb_label ~= k) = 0;
    segmented_images{k} = color;
end
C1 = segmented_images{1};
C2 = segmented_images{2};
C3 = segmented_images{3};
%imshow(C1), title('objects in cluster 1');
%imshow(C2), title('objects in cluster 2');
%imshow(C3), title('objects in cluster 3');
%start example mean_cluster_value = mean(cluster_center,2);
[tmp, idx] = sort(mean_cluster_value);
blue_cluster_num = idx(1);
L = lab_he(:,:,1);
blue_idx = find(pixel_labels == blue_cluster_num);
L_blue = L(blue_idx);
is_light_blue = im2bw(L_blue,graythresh(L_blue));
nuclei_labels = repmat(uint8(0),[nrows ncols]);
nuclei_labels(blue_idx(is_light_blue==false)) = 1;
nuclei_labels = repmat(nuclei_labels,[1 1 3]);
blue_nuclei = he;
blue_nuclei(nuclei_labels ~= 1) = 0;
%end example imshow(blue_nuclei), title('blue nuclei');

From start example to end example, it is supposed to only include the blue part of the cluster. How can I change it to only include the green part?

[Walter Roberson]

A couple of sample images would help.

Answer 1

[Walter Roberson]

So experiment with

mean_cluster_value = mean(cluster_center * [1 0;0 -1],2);

Answer 2

[Image Analyst]

Well, that's not going to work. Not sure why you thought it would. Have you actually looked at the projection onto the a-b plane? Using kmeans is not good - you'd be better off using rgb2ind() (which uses a more sensible color quantization routine) or using fixed thresholds. Even better would be to use fixed thresholds in HSV color space. And why did you think there would be three color clusters in your gamut? Depending on what's in the background, there could be 2 or 4 or 10 or 20.

It's late tonight. Perhaps tomorrow I can explain further and show you the color gamut. In the mean time, try the color segmentation app if you have the last version or two of MATLAB.

[John Conson]

For the most part, Kmeans did an OK job at segmenting the layers (and separating them accurately.) Color segmentation would always get something else into the shot and I am not entirely sure how to utilize it (I am a beginner). Plus, with the amount of different shots I will be processing (150-500), the contrast and overall color of the leaf/ photo may change. Color gamut does sound useful

[John Conson]

How would I exactly segment the image once it is indexed?

[Walter Roberson]

Do you need to do something with the segmented image beyond setting the non-leaf parts to black?

[John Conson]

I have to set a histiogram for each layer of intensity (red, blue, green, etc.) after I set the non-leave part to black

[Image Analyst]

Try this:

% Extract the individual red, green, and blue color channels.
redChannel = rgbImage(:, :, 1);
greenChannel = rgbImage(:, :, 2);
blueChannel = rgbImage(:, :, 3);

[redCounts, redLocations] = imhist(redChannel(mask));
[greenCounts, greenLocations] = imhist(greenChannel(mask));
[blueCounts, blueLocations] = imhist(blueChannel(mask));

Also make sure you check out my lengthy discussion of gamuts and color segmentation in your duplicate question.

[John Conson]

Thanks again for the suggestion. Do you know any other available segmentation options (such as median cut?)

[Walter Roberson]

How much longer until the sample images arrive?

[Image Analyst]

There are lots of color segmentation algorithms. For example, here's one: http://www.mathworks.com/matlabcentral/fileexchange/37197-dem--diffused-expectation-maximisation-for-image-segmentation