Digital flow for shape decomposition: Application to 3-D microtomographic images of snow

• A novel framework for shape decomposition based on digital flow is proposed.
• A fast computation of flow and critical points is presented in the digital domain.
• Two simple thresholds are adopted to control the quality of decomposition.
• The application to various types of 3-D snow images is efficient and precise.

We propose a fast shape decomposition method for granular microstructures using a 3-D approach based on medial axis. We define a two-step algorithm: the first step relies on a notion of digital flow to obtain a preliminary over-decomposition from medial balls. During a second step, we use geometric criteria to obtain a relevant and precise volumetric decomposition. We apply our algorithm to 3-D objects of materials and, more precisely, to microtomographic images of snow microstructures.