How can I Implement Maker-Controlled Segmentation to Separate and Count the Individual Circles in this Microscope Image?

I am trying to implement maker-controlled segmentation (following the code present at https://www.mathworks.com/help/images/marker-controlled-watershed-segmentation.html) to separate the individual circles in this image and then using regionprops to locate the individual circles as centroids. But with the resolution and the shear number of circles, I am unable to effectively separate them. If anyone has any ideas or potential approaches it would be greatly appreciated! I have attempted to use unsupervised ML algorithms like K-means clustering to clear up the background and foreground but I still cannot get clear circles to separate effectively for regionprops to count it.

Thank you so much in advance for your assistance!

[Rahul]

I am not able to visualize the image even after importing in MATLAB. Please share the proper image and the expected output.

Answer 1

[Sourabh]

Hey Christopher,

The sample image you have attached is just a blank image so I could not run tests on it.

However, you might find the following links useful as they tackle similar problems and might be able to help you find an alternative to the maker-controlled watershed segmentation method:

• https://www.mathworks.com/help/images/detect-and-measure-circular-objects-in-an-image.html
• https://www.mathworks.com/help/thingspeak/detect-and-measure-circular-objects-in-an-image.html
• https://www.mathworks.com/matlabcentral/fileexchange/34365-circle-finder

MATLAB also an inbuilt function called imfindcircles to detect circles in an image.

All the best!