<p>Rock thin-section images segmentation refers to the process of dividing images of rock or mineral samples, captured under a microscope, into different regions or objects to extract pore data. These regions or objects may represent different rock components, microstructures, or mineral types. Due to the heterogeneity of rocks and the complexity of diagenetic evolution, the pore distribution observed in thin-section casts is scattered, and the pore shapes are irregular. Such complexities hinder the effectiveness of many conventional semantic segmentation techniques, leading to inaccuracies in porosity assessment and pore characterization. To address the aforementioned issues, this paper introduces a multi-stage adaptive Otsu thresholding algorithm tailored for pore segmentation in rock thin-section images. The algorithm processes the original rock cast thin-section image in three main stages: (1) denoising and color space conversion, (2) coarse segmentation using adaptive thresholds in the Hue (H) channel, complemented by segmentation of the Saturation (S) and Value (V) channels with Otsu thresholding, and (3) refinement of the segmentation results through morphological correction using an eight-connected component method. To validate the effectiveness of this method, we selected cast thin-section images of different lithologies from the Sichuan Basin and the Ordos Basin for a series of comparative studies. Comparative studies demonstrate that the algorithm can accurately and efficiently extract pore characteristics from high-resolution images of rock cast thin-sections with varying colors.</p>