masking out unwanted area in series of images

I have 1000 images sequentially extracted from a video. I am making a grid of small windows in all those images. I want to find out average intensity value of each and every window inside any image. But there are some small circular structure, constantly present in present in all 1000 images which are not the part of my average intensity calculation. How do I get rid of those features.

[Walter Roberson]

I see complete and partial circles, but I also see a lot of dots of various shades of gray. Is it only the large circles that are to ignored, or are the dots to be ignored as well? If so, then what should be the intensity cut-off for those to be ignored (since some are so light that they are debateably part of the intensity that you want to measure.)

[navin chandra]

small dots are useful for my analysis. Bigger circles are top view of pillars in a microchannel which I want to exclude from my analysis.

[Image Analyst]

Right, so I think my answer below would help you. Did you even see it or try it?

[navin chandra]

Thanks for the support but I could not understand-

(i) How to get option to click on smallest triangle.

(ii) How we can extrapolate the triangle into honeycomb structure.

Finally, I used freehand circle drawing and then masked out unwanted area.

[Image Analyst]

(i) Try roipolyold() or ginput(3)

(ii) It's just straightforward geometry -- I'm sure any smart engineer could figure it out.

If you want to do it manually with some freehand masking way, then that's also fine.

[Walter Roberson]

The sample image has disappeared :(

Answer 1

[Image Analyst]

Can you have your users click on the smallest triangle? Then you can extrapolate that out to all points in the honeycomb pattern. I think that would be the easiest way.

Otherwise you can try the radon transform to find out which angles the pattern is alone and try to get the pattern from that.

I think using something like thresholding is going to be tough because the spots are open and the same intensity as the background, and the black borders around the spots are not closed.