Bayesian AD coder: Mesh-aware valence coding for multiresolution meshes

The Alliez Desbrun (AD) coder has accomplished the best compression ratios for multiresolution 2-manifold meshes in the last decade. This paper presents a Bayesian AD coder which has better compression ratios in connectivity coding than the original coder, based on a mesh-aware valence coding scheme for multiresolution meshes. In contrast to the original AD coder, which directly encodes a valence for each decimated vertex, our coder indirectly encodes the valence according to its rank in a sorted list with respect to the mesh-aware scores of the possible valences. Experimental results show that the Bayesian AD coder shows an improvement of 8.5–36.2% in connectivity coding compared to the original AD coder despite of the fact that a simple coarse-to-fine step of the mesh-aware valence coding is plugged into the original algorithm.

Symbol distribution in the decimation conquests: (a) the original AD coder, where 0, 1, 2, 3 and 4 represent the null-patch and the valence-6/-5/-4/-3 codes, respectively, and (b) our Bayesian AD coder.Figure optionsDownload high-quality image (103 K)Download as PowerPoint slideHighlights
► We improve the valence coder for multires. mesh (AD coder) with a Bayesian approach.
► Our coder is aware of the geometry and connectivity of the current mesh.
► We indirectly encode a valence w.r.t. the mesh-aware scores of the possible valences.
► Our coder shows a performance improvement of 8.5–36.2% compared to the AD coder.