How can you find circular blobs in an image?

I am trying to find certain parts of an image, and so used the Canny edge detector. Now I want to get rid of all the edges that are not vaguely circular in shape. How can I remove them?

This is a sample image that I am trying to search.

Answer 1

[Image Analyst]

What if the blob is shaped like a C or a G? Is that roughly circular? If so, you'll need to call bwconvhull() first. If not, you can look at the ration of perimeters square to areas

measurements = regionprops(binaryImage, 'Area', 'Perimeter');
allAreas = [measurements.Area];
allPerimeters = [measurements.Perimeter];
circularities = 4*pi*allAreas ./ allPerimeters^2;

round objects will have circularity values more than around 0.8 or so, or whatever value you want to go with.

[AG]

Will this give me the values individually for each blob, or will it give me an average?

[Image Analyst]

circularities is an array with each element being the circularity of the blob having that index. Does that answer your question? Now, how about answering mine?

[AG]

Thank you so much! In answering your question, the blobs may be C's or G's, as long as the blobs are roughly circular.

[Image Analyst]

Then you'll want to call

binaryImage = bwconvhull(binaryImage, 'objects');

to turn the C an G into solid round blobs. Otherwise they're considered wavy stick-like objects and won't have a circularity close to 1.

[AG]

The values I am receiving are all greater than one. Is that a problem?

[Image Analyst]

If the blobs are all round, it might happen that some values are more than 1. If they are stick-like then the values should be low, like around 0 to 0.6 or so. I actually made a mistake and forgot to square the perimeter. I've now corrected it and the correct equation is also below:

circularities = 4*pi*allAreas ./ allPerimeters^2;