Image-based 2D to 3D reconstruction for complex pore structures

Abstract

The morphology of a porous medium strongly influences its flow behaviours. 3D reconstructions using 2D thin sections enable detailed understandings of those pore structures when no suitable 3D image is available. This thesis focuses on the development of a 2D to 3D reconstruction methodology for complex pore structures. Through a detailed literature review, major limitations were identified for the state-of-the-art reconstruction algorithms, including pattern reproduction, 3D continuity, and computational efficiency, which serve as guidelines for this research. First, a new orthogonal patch-based algorithm was developed, enabling the integration and reproduction of three-directional 2D training images. Second, to further improve the 3D continuity and eliminate error accumulation, an enhanced iterative texture optimization-based algorithm was developed. In this method, a distance map transformation was introduced to enrich spatial information of binary training images, which improves the reproduction of pore/solid size distribution. The issues of blurriness and smoothed pore boundaries in conventional algorithms were analysed and alleviated by improvements in both Search and Optimize steps of the iterative method, which also increase the efficiency. A wide variety of microstructures were reconstructed that closely matched corresponding 3D references in terms of morphology and petrophysical properties, thus validating the quality and capability of the final iterative method. It was then successfully applied to model 3D structures from high-resolution large-scale thin sections, offering significant understandings of the spatial heterogeneity and flow properties in complex microstructures. Overall, this thesis provides novel methods for fast and accurate characterization of 3D complex pore structures using 2D images.