A markerless facial expression capture and reproduce algorithm

This paper presents a markerless facial expression capture and reproduce algorithm. Firstly, an uniform mesh model is built based on the feature points from ASM (Active Shape Model). It can cover 85 percent of the face. Then, a method to capture facial expression based on the face model is proposed. The optical flow tracking is used to track the feature points with particle filter for stablizing the result. The feature points'displacement can drive the overall mesh to deform as the initial value of the mesh tracking. Finally, the captured expression data is used to drive face models with different methods for the models of different dimensions, and then the facial animation can be reconstructed. Experimental results show that the proposed algorithm can capture facial expressions well, and the animation effect is good when mapping the captured expression to 2D cartoon or 3D virtual face models.