find spots in image

I have an image: http://img5.imageshack.us/img5/3617/clb1.jpg

and I use the code, then I can find the spots.

----------

clear all;

Cimg=imread('CropTest1.jpg','jpg');

Gimg=rgb2gray(Cimg);

BW=Gimg>150;

rp=regionprops(BW,Gimg,'WeightedCentroid');

disp('Centroiding is done.');

figure(1); image(Cimg); axis image; hold on;

n=numel(rp);

for i=1:n

    plot(rp(i).WeightedCentroid(1), rp(i).WeightedCentroid(2), 'r*')

end

----------

the result is this: http://img853.imageshack.us/img853/1299/3h2g.jpg

and I have coordinates of all spots.

then I want to do these

- express Center spot -> (0,0)

- find missing spots (-1,3),(5,1),(1,-5) and output the message: '(-1,3),(5,1),(1,-5) are missing!'

- delete one spot in double spots that use this idea: distance of two spots(well, I seem to be able to do this.)

thank you for your advice and help.

Answer 1

[Image Analyst]

I can't see your images, though I'm in Beijing this week and imageshack.us seems to be another domain in a long list of web sites I've encountered that are banned by the government of China. Nonetheless, I'll try to help as best I can without seeing your images. I'm assuming that you have a 10 by 10 array of possible locations of the circles. You first define a logical array initialized to false that says whether or not you have a circle there. Then set it to true when you find a circle there. Try something like this (untested)

circlePresent = false(10,10);
% Then loop over circles you have found setting the array to true.
centroids = [rp.WeightedCentroid];
[rows, columns] = size(BW);
for k = 1 : length(rp)
  if centroids(k, 2) < rows/10
    row = 1;
  elseif centroid(k, 2) < 2*rows/10
    row = 2;
  elseif centroid(k, 2) < 3*rows/10
    row = 3;
% and so on up to 10.
% Then do the same for columns.
  if centroids(k, 1) < columns/10
    column = 1;
% etc.
% Now set the array to true for this circle
  circlePresent(row, column) = true;
end
% Now find missing elements
[missingRows, missingColumns] = find(~circlePresent);
% Print out all the missing circles array locations.
for k = 1 : length(missingRows)
  fprintf('There is a circle missing at row %d, column %d.\n', ...
  missingRows(k), missingColumns(k));
end

[PJM]

Wow... How do you know without looking images? After some modify your answer, I complete the work. Thanks a lot. :D

[Image Analyst]

I guess because I'm that good. :-)

I'm leaving for Japan tomorrow so I'll be able to access sites again.

[PJM]

Thank you for your continuous attention. :D

If you reply later, I compare your answer with mine.

[ajinkya nikam]

i did not get anything. matlab only tells that CONTINUE ENTERING STATEMENT what should do

[Image Analyst]

I have no idea what you're referring to. Start your own thread, post your own code and image, and I'll see what I can do.

Answer 2

[Elad]

I would try imfindcircles(), (or hough transform to find circles if you don't have the toolbox). You can adjust the sensitivity and edge until you find all the spots. Than you can calculate the average intesity in each circle and have full control on your results.

Here is an example in the image processing blog, using imfindcircles: http://imageprocessingblog.com/detecting-and-counting-objects-with-circular-features/

or at steve on image processing blog: http://blogs.mathworks.com/steve/2012/09/04/detecting-circular-objects-in-images/

[PJM]

Thank you for your help.

I tried imfindcircles function. Despite I adjust sensitivity and edge, but there is still one missing point at (1,-5). Even if I find all points, how can I get WeightedCentroid? :'(

[Elad]

You might be able to get the missing point by changing the ObjectPolarity to 'dark', and to approximate the "Weighted Centroid" by the circles centers.

But Nevermind, I see it worked out for you anyway. :)

[Elad]

[centers, radii] = imfindcircles(clb1,[8 15],'Sensitivity',0.91,'Edge',0.01);

gets all the spots but one.

[centers, radii] = imfindcircles(clb1,[8 15],'Sensitivity',0.9,'Edge',0.01,'ObjectPolarity','dark');

gets the last spot.

[PJM]

Wow! actually, I feel the limit of regionprops when bright backgrounds. I will try this way.

Thank you for your help. :D