Abstract: A spherical Voronoi tessellation and spherical harmonics-based approach is proposed for reconstructing high-quality particle surface from X-ray computed tomography. This approach adopts the spherical coordinate system for particle surface parametrization and utilizes weighted spherical Voronoi tessellation to adaptively adjust the locations of the sampled parametric points. Two formulations of weights for the Voronoi seeds are developed, namely one based on radial distance and one based on curvature. The approach proposed can be applied to the surface discretization of particles represented by implicit functions, such as quadrics or spherical harmonics, and to the reconstruction of particle surfaces that are originally obtained from imaging technical such as X-ray computed tomography. It has two advantages: It can provide surface meshes with any number of nodes; It offers a flexible option to control the distribution of the surface mesh nodes. The shape characterization analysis and discrete element simulation of the reconstructed particles are carried out to analyze and to verify the effectiveness and accuracy of the proposed approach. Analysis results indicate that: the approach can effectively simplify complex surface meshes, reduce particle surface nodes, and preserve shape morphology descriptors; thusly, significantly reduce the time cost of discrete element simulation.