Temporal DCT-Based Compression of 3D Dynamic Meshes

This paper introduces a new compression scheme for 3D dynamic meshes with constant connectivity and time-varying geometry. The proposed approach, referred to as Temporal-DCT encoder (TDCT), combines a piecewise affine prediction scheme with a temporal DCT-based compression of the prediction errors. Experiments show that TDCT achieves up to 70% and 54% lower compression distortions than the GV and MCGV approaches, while outperforming (47% to 95% lower distortions) RT, MPEG-4/AFX-IC, D3DMC and Dynapack techniques. Key-Words: Dynamic 3D mesh compression, DCT, piecewise affine prediction, motion-based segmentation.