The geometric and spatial characteristics of soil pores determine the permeability and water retention of soils, which have important effects on soil functional diversity and ecological restoration. However, in the existing research on soil pores, there is a lack of tools and methods for describing intuitive and quantitative characteristics of pores. To solve this problem, the three-dimensional reconstruction technique based on marching cubes algorithm and ray casting algorithm was used to restore the geometry and spatial distribution of soil pores. The objects of experiment were the CT images of soil pores with a single large pore and different pore densities. The comparison results of the two algorithms showed that the pore structures of different sizes reconstructed by MC algorithm had boundary aliasing and missing phenomena, and the pore volume was also smaller than the actual situation. While the RC algorithm reconstructed the pore contour clearly and the structure was real, the detailed information of each pore structure can be completely represented, and the reconstruction effect was not affected by the porous density of the soil sample. In addition, in order to further evaluate the reconstruction performance of the two methods, three indicators of model quality, rendering speed and algorithm characteristics were used to compare and analyze the experimental results. In order to ensure the comparability of the refactoring indicators, all experiments were carried out by using the VTK graphics development library, based on the Visual Studio 2017 programming platform of the same computer. The results showed that the MC algorithm was only for the reconstruction of the pore surface voxels, so it occupied less memory and had faster drawing speed. However, due to the ambiguity of the algorithm, the quality of the reconstructed pore model was general. The RC algorithm calculated all the voxel points of the sample, so it took up a large memory and drew slowly. However, assigning opacity and light intensity were used to each voxel to synthesize the model, avoiding the disadvantages of the MC algorithm, and it was able to maintain the details of the pore model. In summary, the RC algorithm would be able to provide a more advanced method for the visualization of soil pores, laying a technical foundation for studying soil moisture and nutrient transport and air exchange.