Thresholding of a phase with etching but no significant contrast difference (novice user)

Sample image and/or code

S_CML_101_am_ETD_011.tif (12.9 MB)



The attached image features a secondary electron image from an SEM of a metal alloy. There are three phases visible: the background (matrix) phase (Phase 1), the phase with the darkest contrast (Phase 2) that appears flush with Phase 1 and a third phase that is appearing to come off the surface and haloed (Phase 3). Phase 1 and Phase 3 have very similar contrast.

Analysis goals

I am seeking to find the area fraction of the etched Phase 3. I am pretty new to the advanced image processing options but I have attempted a few things:

I have attempted different combinations of image processing (shadows, contrast, etc) and filtering (both gaussian and kuwahara) along with ‘find edges’ but can never quite get the Particle Analysis function to actually fill in the Phase 3 outlines. I was basing this off this approach.

I also attempted Weka Segmentation that got me pretty close when I included Kuwahara filtering but it can’t quite figure out where the Phase 3 borders stop. I have attempted some pre processing before this step but the end result hasn’t been sufficiently satisfactory.

Thank you for your assistance. Any suggestions are helpful!