Abstract Pore and grain phases separation significantly affect the rock physical properties extracted by digital rock measurement techniques (DRMTs). Previous studies mainly focus on approaches to predict the physical properties of rocks, and less attention is paid on the effects of segmentation techniques and voxel resolution on rock physical properties. Physical properties of rocks are not still well understood using DRMTs with less experiment effort. In this work, 6 segmentation algorithms and 4 voxel resolutions combined with X-ray CT imaging are applied to the Stainton and Clashach sandstone, Indiana limestone and Estaillades carbonate samples to understand their effects on rock physical properties. Moreover, threshold-porosity relations are proposed to accurately segment the pores and grains in rocks. Results show that voxel resolution has little to no impact on the rock physical properties. However, the segmentation algorithms significantly affect these properties, in which the Ostu's method, Iterative self-organizing data (IsoData) method and Fuzzy c-mean clustering (FMC) method should be alternative methods to separate the pore and grain phases in sandstones. For carbonates, under-segmentation is still inevitable with these techniques. Rock physical properties calculated by the proposed threshold-porosity relations are closer to the experiment results than those calculated by the classical segmentation methods.

中文翻译：

---

X射线成像对岩石物理性质的分割技术和体素分辨率的敏感性分析

摘要 孔隙和颗粒相分离显着影响数字岩石测量技术（DRMTs）提取的岩石物理特性。以往的研究主要集中在预测岩石物理性质的方法上，较少关注分割技术和体素分辨率对岩石物理性质的影响。使用 DRMT 以较少的实验工作仍然不能很好地理解岩石的物理特性。在这项工作中，将 6 种分割算法和 4 种体素分辨率结合 X 射线 CT 成像应用于 Stainton 和 Clashach 砂岩、印第安纳石灰岩和 Estaillades 碳酸盐样品，以了解它们对岩石物理性质的影响。此外，提出了阈值-孔隙度关系来准确分割岩石中的孔隙和颗粒。结果表明，体素分辨率对岩石物理特性几乎没有影响。然而，分割算法显着影响这些特性，其中 Ostu 方法、迭代自组织数据 (IsoData) 方法和模糊 c 均值聚类 (FMC) 方法应该是分离砂岩中孔隙和颗粒相的替代方法。对于碳酸盐，使用这些技术仍然不可避免地会出现欠分割。由提出的阈值-孔隙度关系计算的岩石物理性质比经典分割方法计算的岩石物理性质更接近实验结果。迭代自组织数据（IsoData）方法和模糊 c 均值聚类（FMC）方法应该是分离砂岩孔隙和颗粒相的替代方法。对于碳酸盐，使用这些技术仍然不可避免地会出现欠分割。由提出的阈值-孔隙度关系计算的岩石物理性质比经典分割方法计算的岩石物理性质更接近实验结果。迭代自组织数据（IsoData）方法和模糊 c 均值聚类（FMC）方法应该是分离砂岩孔隙和颗粒相的替代方法。对于碳酸盐，使用这些技术仍然不可避免地会出现欠分割。由提出的阈值-孔隙度关系计算的岩石物理性质比经典分割方法计算的岩石物理性质更接近实验结果。