Well I'm already sort of doing binary thresholding, It works fine so far, but the issue is more about flexibility if it could work on a lot of different images which isn't the case. I can provide sample images but I will want to be looking at a huge variety of sample images so I don't know how helpful it would be to show a few since other images could be hugely different from the sample ones. The images are from an Atomic Force Microscope, it produces a topography of the sample surface you wish to scan, may be a biological sample (DNA/Chromosomes/Collagen/Enamel), or a chemical compound or silicon/graphene, so Im not looking at any one particular set of images, more interested in a robust solution for any and all images. The objects that appear in the image may be circular in some cases, or elliptical, irregular, lines, e.t.c.
