I have a image of two circles which are partially occulded. I have to label them as two different label. is there any way to label them differently? I am adding some more images. Images have some regular & non-regular shapes.

>>

>>

[poonam]

Hi harjeet, I am adding some more images. My Images have some regular and some non-regular shapes. regular shapes include perfect circle and distorted at the borders. non regular shape are toching to the regular shapes.

>>

>>

[Image Analyst]

poonam, did you see my answer at all????

Answer 1

[Image Analyst]

poonam, did you see the example in the help for the watershed function? This is exactly the help demo - two overlapping circles, separating and labeling them. Just follow the code in the help example.

[poonam]

No sir, I did not knew about it. But I will learn about it. If you can help me more on this I will thankful to you.

[Image Analyst]

Harjeet gave you code. If you have questions about it, ask. Here are some links for you to learn from:

http://cmm.ensmp.fr/~beucher/wtshed.html

http://www.mathworks.com/help/images/examples/marker-controlled-watershed-segmentation.html

http://blogs.mathworks.com/steve/2013/11/19/watershed-transform-question-from-tech-support/

[poonam]

Yes sir, I learned about that function. and it worked very well for all my images. And I think my revision of question was unnecessary. Thank you.

Answer 2

[harjeet singh]

hello poonam do use this code

clear all
close all
clc

img=imread('rgbCattach.png');
img=rgb2gray(img);
figure(1)
imshow(img)
drawnow
img=logical(img);

[centers, radii, metric] = imfindcircles(img,[30 100]);

img_labeled(1:size(img,1),1:size(img,2))=0;
for i=1:size(centers,1)
    for j=1:size(img,1)
        for k=1:size(img,2)
            d=sqrt( (j-centers(i,2)).^2 + (k-centers(i,1)).^2);
            if(d<=radii(i))
                img_labeled(j,k)=i;
            end
        end
    end
end

figure(2)
imshow(img_labeled)
drawnow

[poonam]

hi harjeet, your code working great for this image. But i have many images (other than this) which are not the circle regions and attached same way like this. then i will not be able to use 'imfindcircles' function. then how to separate that two region?

[harjeet singh]

for that i have to see the area by which the objects attached, kindly upload 2 3 images

[Guillaume]

While Harjeet's method is good, the implementation can be improved. This is a more efficient way (less loops) to generate the labeled images:

%follows the line
%[centers, radii, metric] = imfindcircles(img,[30 100]);
img_labeled = zeros(size(img));
[pixrows, pixcols] = ndgrid(1:size(img, 1), 1:size(img, 2));
for circleidx = 1:size(centers, 1)
   pixdistance = (pixrows-centers(circleidx, 2)).^2 + (pixcols - centers(circleidx, 1)).^2;
   img_labeled(pixdistance <= radii(circleidx)) = circleidx;
end

[harjeet singh]

yes its shorter thanks @guillaume

[harjeet singh]

hello poonam as @image analyst told you to use watersed function, that's the best in your case use this code

clear all
close all
warning off
clc

img=imread('s4.png');
img=logical(img(:,:,1));
figure(1)
imshow(img)
drawnow

D = bwdist(~img);
D = -D;
D(~D) = -Inf;

figure(2)
imshow(D,[],'InitialMagnification','fit')
title('Distance transform of ~bw')

L = double(watershed(D));
rgb = label2rgb(L,'jet',[.5 .5 .5]);
figure(3)
imshow(rgb,'InitialMagnification','fit')
title('Watershed transform of D')

L(L==1)=0;
figure(4)
subplot(1,2,1)
imshow(L)
title('bw labeled image')
subplot(1,2,2)
imshow(rgb)
title('jet image')

[poonam]

Thank you harjeet, 'watershed' function suggested by @Image Analysis worked perfectly well in my problem. Now I can extract these object individual.