Advanced framework for an error-resilient parameter analysis-synthesis system of facial animation

A novel framework for the Error-resilient Parameter Analysis-Synthesize System (ErPASS) is proposed for facial animation. Two models named FDP/FAP based facial animation model and the OFP based facial animation model are developed and combined to satisfy different animation quality demands. Optical flow estimation based techniques are added into the system to detect the key animation parameters, describe facial motion, and achieve expression morphing. A hybrid data-compression issue is discussed and embed into the encoder. Some effective error concealment algorithms are applied to improve the reconstructed face animation in synthesizer. The ErPASS can be widely implemented in the very low bit-rate wired or wireless communication applications.