An interpretable morphostructural descriptor for pore architecture in segmented geological images

Pore structure controls fluid storage, connectivity, and transport in geological materials, yet conventional image-based descriptions often reduce pore systems to a single areal fraction. Such simplification may obscure differences in connectivity, fragmentation, heterogeneity, and scale-dependent organization that are essential for interpreting porous rocks. Here, an interpretable (Marion-based) morphostructural descriptor is introduced for segmented two-dimensional geological images represented as binary pore–matrix maps. The descriptor combines apparent two-dimensional porosity, largest-cluster dominance, cluster fragmentation, pore-shape irregularity, lacunarity, skeleton density, box-counting fractal dimension, and departure from a compact 1/10 areal reference. Applied as a proof of concept to three representative fluid-impregnated petrographic sections, the framework separates a compact, fragmented pore structure, a transitional, partially connected pore structure, and a network-dominated pore structure. This separation is not governed by porosity alone, but by the coupled evolution of pore abundance, connected-network dominance, spatial intermittency, cluster hierarchy, and scale-dependent occupation. Multiscale tests with local analysis windows show that the main interpretive rankings remain stable across observation scales, supporting the descriptor’s robustness. The proposed framework does not provide absolute permeability or volumetric storage estimates. Instead, it offers transparent image-based proxies and auditable structural descriptors for comparative petrography, digital rock analysis, pore-network screening, and geological interpretation of segmented porous media. In practice, this provides a low-cost, reproducible screening tool that can flag structurally distinct pore architectures directly from segmented images, ahead of, or as a complement to, more expensive three-dimensional imaging or petrophysical characterization.