The determination of the gray threshold is crucial for the quantitative characterization of digital images. In this study, a new algorithm is proposed to determine the optimal image segmentation threshold. This algorithm is based on a combined analysis of the gray distribution curve and its second differential distribution curve of the digital image. Then, the feasibility and accuracy of the Liu–Cao algorithm are compared with the other 16 algorithms and verified by mercury injection method (MIP) results. Results show that the proposed segmentation threshold algorithm can effectively segment the digital images obtained by various imaging techniques (SEM, CT, FIB/SEM, etc.) and can accurately extract the pore (crack) structures from the image. Image filtering has a certain influence on the gray threshold determination and quantitative characterization, and the impact depends on the quality of the original image. This algorithm can be easily understood and mastered by the researchers and can be widely used in geotechnical and geological areas.

灰色阈值的确定对于数字图像的定量表征至关重要。在这项研究中，提出了一种确定最佳图像分割阈值的新​​算法。该算法基于对数字图像的灰度分布曲线及其二次微分分布曲线的组合分析。然后，将Liu-Cao算法的可行性和准确性与其他16种算法进行了比较，并通过注汞法（MIP）进行了验证。结果表明，提出的分割阈值算法可以有效地分割通过各种成像技术（SEM，CT，FIB / SEM等）获得的数字图像，并可以从图像中准确地提取孔（裂缝）结构。图像过滤对灰度阈值的确定和定量表征有一定影响，其影响取决于原始图像的质量。研究人员可以轻松地理解和掌握该算法，并且可以广泛应用于岩土和地质领域。