Morphological Persistence from SEM: Threshold–Fractal Signatures with Occupancy-Controlled Binarization

Abstract

Scanning electron microscopy (SEM) micrographs are routinely interpreted qualitatively, yet they encode reproducible lateral morphology patterns that can be quantified without invoking height metrology. Here we propose a compact, practical protocol to extract a morphological persistence signature from SEM intensity fields using threshold-dependent box-counting fractal analysis with explicit preprocessing-sensitivity checks. We show that fixed normalized thresholds can fail for certain microstructures by driving the binary representation into a low-support regime at high thresholds, rendering the excursion sets visually uninformative and fractal estimates unstable. To address this failure mode, we introduce occupancy-controlled (percentile-based) thresholding that enforces comparable foreground occupancy across the threshold series, enabling meaningful multiscale comparison. Applying this protocol to two contrasting SEM textures (grain-like and agglomerate-like), we report threshold-dependent fractal dimensions Df(T) and quantify sensitivity to three standard preprocessing strategies (linear rescaling, histogram equalization, and CLAHE). The resulting persistence profiles provide a reproducible SEM-only criterion to compare microstructures and to flag when apparent “fractal complexity” is dominated by segmentation sensitivity rather than persistent image-field morphology.

Keywords: SEM; Image Analysis; Box-counting; Fractal Dimension; Thresholding; Preprocessing Sensitivity; Microstructure; Morphological Persistence

Citation

Vasconcelos HC, Meirelles MG. Morphological Persistence from SEM: Threshold–Fractal Signatures with Occupancy-Controlled Binarization. WebLog J Materials Sci Eng. wjmse.2026.f3005. https://doi.org/10.5281/zenodo.21431094