In recent years, image processing software was wisely applied to identify and analyze pore structure. However, these softwares, such as Photoshop and Image J, did not take into account the complexity of the internal material in the soil and the irregularity of pore structure, and they caused low pore segmentation precision. In order to solve the problem, a pore quantitative method based on the characteristics of soil computed tomography (CT) image was proposed. This method mainly included image processing and quantification analysis. Firstly, the adaptive median filtering algorithm was adopted to remove the effect of image noise on the edge of pore. Then, the method of iterative optimal threshold and canny edge detection was used to identify the pore structure in the soil and the contour line of the pore. Secondly, soil pore structure had evident spatial characteristics, which included soil porosity, pore number, pore radius, spore size distribution, circularity, fractal dimension, and so on. These pore geometry indicators were calculated by using the mathematical statistics method, and they could reveal the complexity and irregularity of pore structure. These geometry indicators were useful for realizing the quantitative analysis of the soil porosity. Finally, the method was applied to the soil under freeze-thaw cycle. The results showed that the method can accurately locate the pore profile, and segment the pore structure effectively. Furthermore, the effect of freezing and thawing cycles on the soil structure was revealed by quantifying the geometrical parameters of various soil pores, thus it proved the effectiveness of the method and laied foundation for quantification of soil pore geometry and spatial characteristics.