Removing Out of Focus Pixels in an Image

Hello

I am currently working on an experiment whereby a high speed camera takes pictures of bubbles departing from the surface. The image taken has depth and thus have some bubbles in focus while some are not. We are trying to process the image such that only the in focus bubbles remain. We are trying to measure diameter, departure velocity, and frequency so only the bubble boundary is important. I have posted some images here:

picture

The main problem I face is that there is overlapping bubbles and the out of focus bubbles have the edges in focus. I tried using the fmeasure function and changing the out of focus pixels to white but ended up deleting the in focus bubbles as well (for some reason the in focus bubbles have a very low focus measure) while retaining some out of focus bubbles' boundaries.

Any help is appreciated. Thanks

Answer 1

[Image Analyst]

I think the magnification is way too high to do good tracking. I think you should reduce the magnification. I don't think you need to worry about the boundary. I would think that getting the velocity from the centroid of the bubble would match pretty darn well with any velocity you would derive from finding the boundary. And I think that you'd have an easier time tracking individual bubbles with lower mag. Or you can talk to a good microscopist who might be able to recommend a better imaging method, such as Cytoviva (see the complex emulsion video here: http://cytoviva.com/gallery.htm) It's always easier if you start with a good image than to try to fix up a bad image. Bottom line, this image is too difficult to analyze automatically. You could get faster success if you allowed some manual assistance such as manual outlining or tagging of centroids or something.

[Lu Bin]

Hi, thanks very much for the response

Unfortunately due to experimental constraints there is a limit to how much we can zoom out (we are looking at formation in a specific location), but we will try to get a larger field of view.

The reason I want to look at the boundary is because we want to calculate bubble diameter as well, for which we will need to calculate the area. If analyzing this automatically is indeed too difficult then we will have to resort to doing this manually.

Thanks very much

[Image Analyst]

Well I can't think of any way. There is just way too much background clutter that is the same darkness as the foreground bubbles and blends in with them. So with irregular shaped boundaries that overlap and combine with other bubbles, it looks very challenging to say the least. I think you'd spend more time trying to develop an algorithm for it (think Ph.D. project) than you'd save by just doing it manually. So, is it more important to get the job done and get the measurements, or is the algorithm development the main task? Are you an imaging scientist . . . or a material scientist or chemist?