Coupling Octree Cross-Correlation Simulation and Watershed Algorithm for Multi-Scale Reconstruction of Rock Images

Abstract

Summary Digital Rock Physics (DRP) serves as a versatile tool for modeling various rock types. However, the complexity of modeling heterogeneous rock samples presents a significant challenge because of their complex structures. Striking a balance between high-resolution (HR) images and a wide field of view (FoV) is crucial for accurate pore-scale modeling, yet the trade-off between resolution and FoV complicates this task. Moslemipour et al. (2023) introduced the Cross-Correlation-based Simulation (CCSIM) with Watershed-based Algorithm (CWMR) for multiscale image reconstruction, although computational costs still remain a concern. This study introduces an innovative approach addressing computational limitations while maintaining reconstruction accuracy. Unlike CWMR, the proposed approach utilizes the Octree structure for µCT image reconstruction of a Berea sandstone. The process involves converting low-resolution (LR) images into an Octree structure, with calculations performed on resampled LR images. This streamlined approach reduces time-consuming computations and minimizes memory consumption. Comparative analysis against new approach and HR images demonstrates the new approach's efficacy, showing close agreement with HR image. The introduced approach completes reconstructions 1.8 times faster (68 hours vs. 120 hours) and requires significantly less memory (2.4 GB vs. 65 GB) than CWMR, highlighting its practical advantages in the realm of DRP.