三维几何模型的L₀去噪算法

It is a fundamental problem to denoise 3D shapes (e.g., triangular meshes and point clouds) in computer graphics. For a noisy mesh, its local region is smooth, or contains geometric features. This paper introduces a shape denoising algorithm based on L₀ sparse optimization. First, we present a L₀ sparse constraint with respect to position and normal of mesh vertices, which can distinguish features from noises effectively. Second, we deal with L₀ optimization problem using an improved optimization scheme, which exhibits better convergence. Furthermore, we extend the above algorithm to denoise point clouds. Massive experiments show that our algorithm can handle high noises, recover geometric features, and obtain satisfactory results.