Research In Motion Capture And Cloning Of Human Facial Expression

Facial expression capture is a branch of motion capture. In various research of related areas, due to the precision, efficiency, convenience to handle and maintain, optical motion capture analysis system is widely used, and studied. Recently, the application of motion capture has involved into motion analysis, Game, TV programs, and many other areas. And demanding of life like facial expression is enlarging dramatically. Expression cloning is an extension area which is essential to facial expression motion capture. Currently, most methods to realize the idea require expensive equipment such as Cyberware and lot of manual work. In order to extend its convenience and application area, this paper do a lot research on expression cloning and 3D facial animation based on markerless monocular motion capture.In Motion capture module, this paper realizes semiautomatic motion data extraction. Using methods in pattern recognition and image processing area, we captured and tracked 18 feature points in actor's face. Tracking result of 18 feature points is encouraging after only a few steps of manual initial work. Motion data will be saved in our defined data structure after normalized.In the face model reconstruction module, a frontal and a profile photo is required. After manual mark feature points on the two photos, 3D model can be created in 10 minutes. All the models should have same amount of vertex and faces. Our model is saved in OBJ format. In order to driven model, we design Elastomer metamorphosing models for eyes and mouth area. So when load motion data in to feature points, surrounding points will move respectively with feature points.The last module is expression cloning, which intend to transfer expression from template model into target model. To realize the transplation, we use texture morphing to calculate relationship between textures of template model and target model. Also we process the projection of texture in order to driven 3D Model. All the target models and template models should be pre-processed before cloning expression.This paper introduces and applies many methods, such as SNAKE used in detecting lips and morphing in calculate the relationship between textures. The equipment requirement of our system is only a camera and a common PC...