Abstract: To reduce the storage size and improve the rendering efficiency of multi-resolution sparse voxels, a wavelet based compression and rendering algorithm is proposed. At the building stage of sparse voxels, according to the multi-resolution characteristic of wavelet and the sparsity of voxel structure, high-resolution sparse voxels were transformed into low-resolution sparse voxels and multi-level detail information by employing 3D Haar wavelet transform, and the wavelet coefficients were encoded with a compact encoding method. At the interactive rendering phase, in order to implement multi-resolution rendering, the low-resolution voxels were selected as shading primitives during the interaction. After the interaction process, the details were added to the coarsegrained voxels level by level through the inverse transform of 3D Haar wavelet to restore high-resolution voxels. Lastly, the rendering algorithm was accelerated in parallel by utilizing multi-core CPU. The experimental results show that the proposed algorithm provides an efficient and achievable way to render models with various complexity.