Comparative analysis of microstructural segmentation techniques on SEM and optical micrographs

Since grain morphology, distribution, and interfaces have significant effects on mechanical and functional properties, accurate microstructural characterization is still indispensable to materials research. The segmentation of Al+SiC nanocomposites using backscattered electron SEM images and the segmentation of Al-Zn-Mg powder metallurgy alloys using optical micrographs are compared in this paper. This comparison, rather than reporting material-specific results, attempts to present a broader methodological perspective by trying to distill general insights into how various imaging modalities drive algorithmic choices. These accuracy aspects are not revisited here because both have already shown high accuracy in automated segmentation and grain boundary detection. Rather, focus is on how image modality, preprocessing specifications, segmentation pipelines, and post-processing techniques interact to determine the robustness, flexibility, and possible transferability of automated microstructural analysis workflows.