Drop size distribution and surface coverage by MATLAB

Dear image analyst, I am doing drop condensation experiment on superhydrophobic surfaces. I would like to get plot total number of drops vs. drop diameter curve (an example of such plot is attached). Also, I would like to calculate surface coverage (i.e., ratio of area covered by drop to total area). Can anyone help me please to develope the code and complete the atatched m file. So far, I can not get a perfect binary image to detect circles. Is there any way to improve the binary image from this original image? I shall highly appreciate your cooperation.

[JUBAIR AHMED SHAMIM]

I have added a better image for analysis. If someone can help me to develop the code for dropsize dristribution I would highly appreciate it. Thanks for your cooperation.

Answer 1

[Image Analyst]

That's a tough one. I'd try to see if you can get a better image. Try different lighting, different geometry, different background, etc. See if you can have different density so that the bubbles don't overlap. Maybe try polarizers. Maybe you can send the fluid through a small transparent tube so that there is just one bubble at a time passing through and then compute the volume between the menisci.

[JUBAIR AHMED SHAMIM]

Dear Image Analyst, many thanks for your message and useful suggestions. I agree with you. I will try a polarizer first in the next week and see if I can get a better image. But this time we have to deal with multiple drops at the same time. Actually, getting drop size distribution is not a problem if I can get a good binary image. Right now if I convert to binary, I can't get a good one. Anyway I will get back with updates after using polarized lenses.

[Image Analyst]

Well of course. Once you have a binary image it's trivial:

props = regionprops(mask, 'EquivDiameter');

but the problem is getting the binary image. I don't see an easy solution but you can search the literature:

https://www.google.com/search?hl=En&as_sitesearch=www.visionbib.com%2F&bthG=Google%2BSearch&q=bubble

for published papers on bubble detection.