Which mask can I use to count particles in an image?

Hello, I have this image and I want to count particles in it. I have binarized the image with the code reported below and now I have to count particles. It was told me that I have to find a mask that compare white particles to an objet (maybe a circle) of established dimensions and if the object is contained in the white particles than I can consider it a particle, if is not I go on. What can I use?

clear all
close all
clc
I=imread('dapi_cd105-FGF.jpg');
I2=I(:,:,3);
BW=imbinarize(I2);

[Max Murphy]

There are some good starter suggestions in this answer.

Answer 1

[Image Analyst]

There are a number of measures you can do to compare it to a circle, such as the aspect ratio (ask for BoundingBox), Solidity, and circularity (ask for Area and perimeter);

props = regionprops(mask, 'Area', 'Perimeter', 'Solidity', 'BoundingBox');
allAreas = [props.Area];
allPerim = [props.Perimeter];
circularities = allPerim .^2  ./ (4 * pi * allAreas);
bb = vertcat(props.BoundingBox);
aspectRatios = bb(:, 3) ./ bb(:, 4);

Aspect ratios, solidities, and circularities should all be around 1 or so. Anything more than about 2 or 3 is not very circular.

[Marianna Pizzo]

Thank you so much, finding areas I maneged to get the number of particles.

[Image Analyst]

If all you needed was a count of the blobs in the binary image, you could use bwlabel():

[labeledImage, numberOfBlobs] = bwlabel(binaryImage);

but that's on the original image, not the number that you'd have after you post process to extract only round blobs.

Answer 2

[KALYAN ACHARJYA]

No need any special mask, use regionprops function

Steps:

1. Convert to binary image with adjustable threshold value
2. Counts the blobs in binary image, here blobs represents the particles in the image

As first step you have already done, for second step here is the code

blobs_data=regionprops(binary_image);
num_blobs=numel(blobs_data);

Or

From your code

I=imread('dapi_cd105-FGF.jpg');
I2=I(:,:,3);
BW=imbinarize(I2); 
blobs_data=regionprops(BW);
num_blobs=numel(blobs_data)