I am processing some videos to obtain the trajectories of some droplets, and have tried different algorithms stitched together, which work pretty well for most frames, but not on all. I then stumbled upon the Radon transform, and if I understand it correctly it is possible to obtain circles from the inverse Radon transform if the transform is a horizontal line covering the whole range. That means that once I have the transform of one of the frames, I must extract a horizontal line joining the points of highest values on opposing sides of the transformed image. This is where I got stuck, since I tried binarizing the image with different thresholds and other transforms, but found no straightforward way of getting the lines I was expecting (red lines in the attachment). What I would like to get is a binary image of these two horizontal lines, so that when I do the inverse radon transform I get the circles I am looking for, any ideas on how to go about this?
Attachment: rtwanted.jpg, URL: ,