I want to count bright spots in a region on an image. How to do it ?

There are images from microscope in which I need to count number of bright spots at some bright patches. Pls help.

[Rik]

What have you tried already? My first thought would be to try some form of edge detection to see if it is easy on such an image to define a threshold.

[bathery]

I had tried to threshold the binary image but I find problems - The central patch is brighter than many dots outside. Thus a simple threshold algorithm sees large patches at the central region.

My objective is to find (1) the bright dots outside this central bright patch and (2) the brighter dots seen on this central bright patch.

Answer 1

[Image Analyst]

Try taking the red or blue channel, which ever has more contrast. if you want the small, compact bright spots, then try using a top hat filter, imtophat().

I tried to do something with it, but it's a JPG image so the jpg block artifacts are very severe - so severe that they basically prevent image analysis of this image. Please post the non-lossy image, like a PNG, BMP, or TIFF version.

Answer 2

[Bshara Murr]

I = imread('microscope.jpg');

gray = rgb2gray(I);

binaryImage = gray > 110;

imshow(binaryImage);

figure; imshow(I);

[labeledImage, numberOfCircles] = bwlabel(binaryImage);

numberOfCircles

I hope it does the job. You can change the threshold value(110 in my code) to what suites you. You can add a slider as well so you won't have to change it from your code.

[bathery]

Thank you for the help but I had tried this way before and did not work. The central bright region appears as large patches giving errors.

My objective is to find (1) the bright dots outside this central bright patch and (2) the brighter dots seen on this central bright patch.

[Can Koyuncu]

You can use hmaxima transformation. Luckly, Matlab has built-in function. Usage is as following: First convert image into gray scale. gray = rgb2gray(im); Then extract regional maxima whose depth is larger than 25. (You can play with the h value) regions = imextendedmax(gray, 25);

Answer 3

[Ryan Comeau]

Hello, another technique that will maybe work is the use of regionprops. What you could do is, have RP return the brightest values on the image and then sort them yourself. If for example you require a specific shape, you can have RP return the Semi Major and semi Minor axis length for each region (look at the documentation, it's quite elaborate). Somthing like this maybe

image=imread('path/to/image');
imBW=imbinarize(image);
qq=regionprops(imBW,image,{'Centroids','MajorAxisLength','MinorAxisLength','MaxIntensity'});
for i=1:length(qq)
    if qq(i).%conditions for bright circles(MajAL==MinAL, or use cicularity)
        data_vault(i,:)=qq(i).Centroids
    end
end

Hope this helps.