Temporal wavelet-based compression for 3D animated models

Abstract

We present an efficient compression scheme for animated sequences of triangular meshes of the same connectivity. The proposed algorithm exploits the temporal coherence of the geometry component by using a temporal wavelet filtering. The quantization of the resulting wavelet coefficients is then optimized by a bit allocation process. This process dispatches the bit budget across the coefficient subbands according to their influence on the quality of the reconstructed sequence for one specific user-given bitrate. The proposed scheme is simple, fast, flexible, and scalable in frame rate and bitrate. Moreover, simulation results show that our approach is competitive for any kind of animated models, whatever the characteristics (parametrically coherent or not, fine/coarse meshes...), contrary to the related works.

Remeshing and spatio-temporal wavelet filtering for 3D animations

“…The figure 6 shows the curve "rate-distortion" of the DOL-PHIN obtained with 3 methods: i) using a temporal wavelet filtering without remeshing (coder of [2]); ii) using a temporal wavelet filtering after remeshing (proposed remesher and coder of [2]); iii) using a spatio-temporal wavelet filtering after remeshing (proposed remesher and proposed coder).…”

“…To create a spatio-temporal decomposition of this semiregular animation, we first apply the temporal wavelet filtering of [2]. This filtering is based on a lifting scheme.…”

“…• then we encode the set of spatial wavelet coefficients ({H s L t r s }) with the zerotree coder of [7] by imposing the optimal bitrate (R S ) computed by the allocation process of [2] for the encoding of the temporal low frequency subband.…”

“…To take into account the spatio-temporal decomposition, we combine the temporal wavelet-based coder of [2] and the zerotree coder of [7] as follow:…”

“…It means that the triangulation is the same at any frame, and so the number of vertices is fixed. This temporal regularity has been already exploited with a monodimensional wavelet filtering [1,2] in compression algorithms. On the other hand, the spatial wavelet filtering has been rarely adressed in this domain [3,4] because of the irregular sampling of the animations, which makes the filtering complex.…”