Face Modeling And Face Animation Based On MPEG-4

Face modeling and face animation is an active research field in computer graphics all the time. There are many extensive applications of face modeling and face animation to a lot of industrial fields, such as film production, computer game, videoconference, human-machine interface, and so on. MPEG-4 is a new generation standard of video compression. Object-oriented coding is introduced and any object in the scene can separately be coded in MPEG-4. Human face is an important object in MPEG-4, Face Definition Parameter (FDP) and Face Animation Parameter (FAP) define the characters and motions of human face.In this thesis face modeling and face animation are studied. In view of face modeling we present an automatic method of the personalization of face model. First, Cyberware laser scanner scans the face of a specific person to acquire its geometry data and texture image. Then, the 3D coordinates of the feature points on the face model can be decided upon the geometry data and texture image of the specific face. A deformation method based on interpolation function is adopted for face model calibration. Last, texture mapping is used to map the texture image on the face model. A customization face model is generated and can be rendered. In view of mimicking the face expressions and motions, a new method of face animation based on MPEG-4 is introduced. First, according to MPEG-4 face model standard, face model can be divided into some function regions. There are some control points related to FAPs in each region. The pseudo-muscle method is used to implement all the low level FAPs which mimic the elastic deformations of the skin in each region and the rigid motions of the face. To generate the face animation freely, we construct a multi-level control model of face animation which is complex to simple, low level to high level from the bottom up.Last, we develop a 3D face animation system compliant to MPEG-4.In our system, FDP stream is imported to personalize the generic face model. The result of personalization can describe the characters of a specific human face. FAP stream is imported to animate the face model. The animation result mimics the motions of facial expressions and motions in real time.