Bubble Image Diameter Analysis

Hi everyone, I’m trying to measure diameter of bubble images but I got a difficulty in image threshold. the gray color is so close to the bubble color even the reflection inside bubbles.
I try to measure it one by one and it takes lots of time. Any solution or suggestion please share :slight_smile:
PS. I used the backlight technique to take this picture if anyone have suggestions on camera works also please guide me.

Thank you :slight_smile:


This is a tough one. It’s really hard to distinguish the ‘bubbles’ from the background… Also - the background varies a lot across the field of view. This is a case where I would go back to the scope (or a different scope) and work on the acquisition side of things - and also perhaps the sample prep? I’m assuming you acquired this on a wide-field scope? How did you prep your bubble samples for imaging? Perhaps asking for more help on the Microforum will be a good place to start regarding addressing acquisition/sample-prep issues/suggestions.

You can also try looking into machine learning tools such as Ilastik or TWS when it comes time to segment. That might help with some of the imaging issues you are seeing. Also consider saving your data as TIFF files - not JPEG. JPEG files corrupt your data… so you want to also start there with better data preservation formats.

1 Like

Thanks for your suggestions. I took this microbubbles picture from a macro lens camera with backlight technique. Microbubbles are generated by feeding air into water in a transparent column. I will ask in the Microforum as well.

1 Like

Neat project! I think @etadobson is spot on, I also would suggest ilastik and avoiding JPG!

You may also be interested in this paper: A simple technique for the automation of bubble size measurements

Hello James,
I will just add my two cents worth, the backlight procedure is the proper way for what you wish to do, however if you diffuse the light some way, such as a piece of paper or plastic between source and bubbles would help. Also try to use a colored light such as a red or green light instead of white. I do not know exactly what properties you are looking for so the colored light will give a better segmentation basis.
It should be very do-able if we know what exactly it is you are trying to aquire as data.
Let us know.

1 Like

Thanks, it’s interesting.

Thank you, I great idea.
I will try it today. I use LED light 5600K 100% Brightness for capturing a plane layer of microbubbles.
I desire to analyze microbubbles properties as bubble diameter, bubble distribution and gas holdup (if it could, some papers claim gas holdup can be calculated from optical method).

I’ll give the feedback on this method, thanks again :slight_smile:

1 Like