As exploration of medium and shallow oil and gas reservoirs progresses to later stages, the focus is shifting towards deep (burial depths > 4500 m) and ultra-deep (>6000 m) reservoirs. These reservoirs experience high stress, significantly affecting the pore structure of rocks and influencing the microscopic flow of oil and gas. Digital rocks are essential for simulating pore-scale flow, yet current reconstruction methods do not adequately account for high-stress effects. This study introduces a novel method for reconstructing digital rocks under high stress using the discrete element method (DEM). The first step involves transforming CT scan results obtained under room temperature and stress into a DEM model. CT scan images are segmented using the watershed algorithm, particle contours are represented with spherical harmonic functions, and these are converted into clump particles in PFC3D. The DEM model is then established with porosity and particle size distribution matching the actual rock, validated using two-point correlation and linear path correlation functions. In the second step, micro-mechanical parameters for the contact constitutive model are calibrated, stress simulation calculations are performed, and results are converted into voxel data. The third step involves analyzing the geometric and topological structure of pores under different stress conditions and examining the evolution of permeability. The feasibility of this digital rock reconstruction method is validated using Bentheim sandstone as a case study. The study results indicate that the application of stress reduces porosity by 6.21% and permeability by 17.61%.