I am processing some videos to extract the trajectories of bubbles, and have had some success in getting information using different transforms with many parameters and then using Closing, Erosion, Dilation, etc., to get an estimate of the position of the bubbles using ComponentMeasurements. However, due to the amount of parameters I am using it work well for many frames, but not all, so I decided to try to obtain the approximate location of these bubbles using transforms without parameters. In particular, using the help from FillingTransform and then thresholding the image I can get the image attached to this post. There is a large bubble, and a smaller one below it, and I would like to get disks that cover them. The problem is that I have been doing it on a trial and error basis using Closing, and Closing with DiskMatrix kernel, and although the results are OK most of the time I think there might be a better way, any ideas on how I can go about getting this mask of covering disks?
Just as a side note, the other smaller elements and the water mark can be taken out after using SelectComponents, so the real issue is getting the disks right.
Attachment: binaryt.jpg, URL: ,