An anisotropic denoising algorithm for mesh models

Mesh denoising is an important tool for geometric processing. Existing methods still suffer from feature blurring or shape degeneration. To address this problem, we present a new anisotropic denoising algorithm which is based on global optimization. Compared with isotropy, anisotropy is more robust to the irregular sampling of 3D meshes. We define a weighting scheme, which contains three factors: area, position, and normal. Moreover, the global optimization problem is solved alternatively and iteratively to obtain denoising result. Various experiments and analysis show the advantages of the proposed algorithm. Our algorithm can remove noises effectively, and preserve mesh features such as sharp edges and corners robustly.