Optimised mesh traversal for dynamic mesh compression

Abstract

During the last decade many algorithms for compressing 3D animations represented by sequences of triangular meshes have been proposed. Most of these algorithms are lossy in their nature, i.e. the reconstructed data do not exactly match the algorithm input. Quite surprisingly, most of the existing algorithms mainly use only general compression techniques, such as entropy coding, quantisation, PCA or wavelet decomposition, while the inherent geometrical properties of the compressed surface remain unexploited. In this paper we focus on geometry specific optimisation: we extend the PCA-based dynamic mesh compression by optimising the order in which the mesh is traversed. By considering the distribution of residuals and optimising the gate selection strategy we achieve data rate reductions by 5.9–29.1% over the existing approaches in the experiments, while the error introduced by compression remains unchanged. This optimisation improves the performance of our encoder above the performance of current state of the art algorithms.