3D Face Modeling from Video

In this paper, a novel 3D face modeling approach from monocular video is proposed. The core of the approach is to deform the generic face model based on the facial feature information extracted from video frames in order to reconstruct a specific 3D face model. The approach composed of following steps: marking facial feature points for the first video frame, and extracting face contour at the beginning, then performing affine correction to track these feature points in the following video frames. At the same time, face contour is extracted. Generic face model is employed to adapt the actual 3D face by deformations based on the facial feature information extracted from video frames. A specific 3D face model is generated at last. Texture mapping is carried out on the 3D face geometric model to generate a realistic 3D model of a specific human face at the final step. Experiments show that this approach is effective and easy to use.