Weighted bit allocation for multiresolution 3D mesh geometry compression

摘要

In this paper, we propose an efficient geometry cmpression method well-adapted to densely sampled semi-regular triangular meshes. Based on multiresolution analysis performed by wavelet transform, it includes a low complexity model-based bit allocation across wavelet subbands. The main contribution of this paper is the resolution of the sub-optimal bit allocation problem related to biorthogonal wavelet coders for 3D triangular meshes. Indeed, using biorthogonal filters weights the MSE distortion of the reconstructed quantized mesh. These weights can be computed from the wavelet filter bank. This permits to obtain a simple but efficient surface adapted distortion criterion for 3D wavelet coefficient coordinates, and to highly improve the perfomances of MSE-based codecs.