The Facial Expression Transfer And The Avatar Facial Expression Systhesis

Face expression synthesis is one of the key technologies in the field of character animation of computer graphics.The research about this technology has existed for decades and it has been widely used in many fields such as film animation,video conferencing,game entertainment and virtual avatar.Virtual avatar is one of the most entertaining and practical applications.Then the main work of this article is about facial expression transfer and facial expression synthesis of virtual characters.This thesis is based on the face detection algorithm of local constrained model and the algorithm of facial expression transfer.Those two algorithms were implemented simultaneously in Unix system.This thesis also implemented the face detection algorithm of the active shape model,and used many face images in the CeleA database to conduct a comparative experiment between ASM and CLM.Then it proved that the feature points are not independent when they are detected,but affect the detection between each other.Therefore,it is strongly proposed that the method of labeling a feature point in the center of the eye is more conducive to the detection of open eyes,while avoiding the detection errors of upper and lower eyelid adhesion.This thesis has implemented the algorithm of facial expression transfer and carried out a sample experiment with the CeleA database and PICS database.It was found that there were completely inconsistent results with the input expression in the area of eyes and mouth under the dual role of face feature detection error and face expression transfer error.This thesis comes up with a method to improve the accuracy by solving the boundary effect of nuclear density estimation.