Difficult Particle Segmentation


Hi everybody,
I’m relatively new to IA, and I would be so grateful for every suggestion or help.
I have a set of SEM images (greyscale 8-bit), like that I reported as an example. In these images two different mineral phases (the withe one and the gray one) are present together with the background (grey). Unfortunately I’m interested in segmenting the grey one (some cristal indicated by arrows as an example), which is furthermore not always constant in the pixel intensity. Particularly I’m interested in extracting a rough particle morphology, and the total particle area.

At now, I did some attempts:
-image contrast and brightness adjusting.
-applied filters like gaussian blur or median in order to homogenize the pixel intensity inside minerals.

It didn’t work very well, since it seems that the background has almost the same range in pixel intensity as the mineral phase, and the difference between background and minerals is their darker rim.
Is there any solution to take advantage of these rims??

hope I was clear to expose the problem,example.tif (2.3 MB)



Hello Pietro,
I guess you need to look into marker-controlled (seeded) watershed with this, may be some additional filtering will be required for good watershed.
Check this recent thread: Creating binary image from SEM image of ductile dimples on fracture surface

Best regards,


Hi Ilya, many thanks for your suggestion!! I wasn’t aware of that, I’ll check for sure.

best regards,