Bubble detection and diameter estimation

Hey,

I wanted to detect and measure the diameter of the bubbles (elliptical and circular) in this image. I tried various techniques but couldn't do so.

Please help me out.

Answer 1

[darova]

• binarize image
• use regionprops with EquivDiameter option

[Avdhesh Vermani]

Could you please share the code.

Actually the image has some reflections and noise in general so, the results I'm getting aren't very satisfactory. So a code would be very helpful.

[darova]

Try imclose to filter your image a bit

Use bwareaopen to remove small undesired object

[Avdhesh Vermani]

Hey, could you suggest some way as how to binarize the image so that the bubbles and the background are segmented separately like I just want the bubbles and nothing else in the background. Or if you could suggest some way to segment bubbles from the background that would also be very helpful.

[darova]

Honestly i can't. You have a hard image. Don't know how to separate bubbles. THe only ideas i had - i said before

Answer 2

[Image Analyst]

Sounds like you got it figured out since you accepted an answer but I'd first have improved your image capture by placing a black velvet shroud behind your vessel to remove background clutter. Then I'd take an image with no bubbles and subtract the subtract that from the image with bubbles using imabsdiff(). Then threshold and call regionprops(). Realize that the bubbles may overlap and there is no way you can split them apart unless you do something like make a real time CT image. But the area fraction or size distribution of what you do have may be enough for you do differentiate different bubble makers or different bubble size distributions.

But anyway, glad you got it figured out.

[Avdhesh Vermani]

Hey, could you suggest some way as how to binarize the image so that the bubbles and the background are segmented separately like I just want the bubbles and nothing else in the background. Or if you could suggest some way to segment bubbles from the background that would also be very helpful.

[Image Analyst]

I did. And I'm waiting for you to upload a new picture with the black background. When can you post it?

[Avdhesh Vermani]

The thing is that the images were taken sometime back and now it is really not possible for me to take new images. I tried segmentation based on texture and it helped me remove some issues but overall it didn't help much.

[Image Analyst]

Maybe it was from a video. Is there some way to get an image of the apparatus before the bubbles started appearking? Like from the first frame of the video?

[Avdhesh Vermani]

I have attached a few images; one of an empty beaker, one of a beaker filled with the solution. I have also attached the full bubble image. Hope this helps.

[Image Analyst]

It might help but you'll need to scale your image by the ratio of the intensities, because you can't subtract two images if they're not the same exposures. And it looks like they're not.. Pick some part of the scene that is not expected to change at all then divide the image by the value there and multiply by the value of the reference image there. Then call imabsdiff(). See if that works. Hopefully everything will subtract away except the bubbles. If that doesn't work you may not be able to do it unless you manually trace the bubbles or snap new images.

[Avdhesh Vermani]

I'll try this but can you tell me how to manually trace the bubbles? Like I need to draw ROI using image segmenter app right, or are there other methods as well?

[Image Analyst]

See attached demos for imfreehand().