Research On 3D Facial Animation Synthesis Based On MPEG-4

Computer simulation of real face has long been a very active research field in computer graphics and human-computer interaction.Life-like facial synthesis is one of the most fundamental and the most difficult problems in computer graphics.As development of computer graphics in modeling,rendering and animation,facial modeling and facial animation has been widely used in many areas,such as film, advertising,teleconferencing,video phones,digital entertainment,human-computer interaction,virtual reality,computer-assisted teaching,and medical diagnosis. Therefore,as the current focus of research in the field of computer vision and computer graphics,facial animation has not only theoretical significance,but also great application value.In the more than thirty years when research of facial animation carried on,many scholars made a number of innovative research and implemented several facial modeling and animation system.However,there are still many areas waiting for further research and improvement,such as complexity of algorithms,ease of 3D data collection,convenience and robustness of realizing variable posture.On basis of current 3D facial animation research,this paper mainly concerns facial animation generation algorithm based on MPEG-4 standard,as well as facial animation synthesis method which can blend multiple movement elements including lip motion and facial expression.Driven by markup text input,realistic facial animation full of variety is synthesized with low complexity under real-time computation.As to animation synthesis based on MPEG-4 standard,this paper studies facial modeling and FAPs(Facial Animation Parameters) data extraction method.This paper generates a facial model compatible with MPEG-4 standard by using modeling tools,converting into VRML format which describes according to organ parts,and setting FDPs(Facial Definition Parameters) manually.For data collection,this paper proposes an algorithm suitable for FAP extraction.Through analyzing displacement of marked points in human face from video sequence,FAPs which reproduce real facial movement can be calculated.As to facial animation synthesis with rich facial movements,based on characteristics of Chinese language pronunciation,this paper establishes 3D mouth shape library described by FAPs sequences.Based on statistical research of JAFFE database,combination model between emotions and expressions is figured out and used to calibrate the original collected expressions,which is used to establish basic facial expressions library.Finally,based on characteristics of FAPs,this paper puts forward a blending model for lip motion and facial expression.Experiments show that this blending model can simulate realistic facial complex non-linear movement, and can also meet real-time computing requirements.