Visual-driven 3D Face Modelling

Creating 3-D human face model and provide direct visual-driven, is a difficult problem in areas of computer science such as Virtual Reality and Human-computer Interaction, because face has much more degrees of freedom than most available input devices provide. In this paper, we present a 3-D human face model and user can control it by acting out poses in front a single video camera. We improve the existed muscle model according to the physical characteristics of muscle, and use it to drive the face model deformation; propose a facial feature triangle method for extract facial motion parameters form video sequences. The experimental results reveal that the proposed method is robust to illumination change and focal length change, can effectively extract face motion parameters from the video.