Partially overlapped grain shape restoration of the granular rock images

This paper introduces a novel method for restoring each grain's original shape from an image of a granular rock sample whose grains partially overlap with each other. Image-based grain shape restoration provides a promising way to understand the generation of the rock that involves sedimentary and diagenesis processes. The difficulties of this task are: 1) how to effectively identify the region of each grain that is partially overlapped with its neighbor grains; and 2) how to recover the overlapped part of a grain based on the residual part. To solve the first problem, a differential erosion strategy that combines concave-points-constraint and pore-skeleton-extension-erosion is proposed to overcome the under-partitioning issue that popularly occurs in the current erosion-dilation grain partitioning method. To address the second problem, closed B-spline interpolation is applied to recover the original shape of the grains based on the partially presented grain edges. The proposed method successfully processed some synthetic and real granular rock images and presented a competent ability to deal with the moderately compressed granular rock samples consisting of relatively high-roundness grains that each grain can be treated as a convex polygon.