3D Surface Filtering Based on Local Parameterization

A new approach for 3D mesh surface filtering is presented. A local neighbor area is first found for each vertex. Then local plane parameterization is applied to the area. After uniform sampling to the parameterized area, each vertex corresponds to a local image patch. Then wavelet filtering is applied to the image patches. After vertex resampling, filtered mesh surface can be reconstructed. The major contribution lies in the idea of universal vertex local filtering. The filtering algorithm can operate nonzero-genus meshes directly and smoothly. Experiments demonstrated potentials of the algorithm.