Using realistic face models and photometric modeling techniques, we present a visual feedback loop that tracks a face, without any marker or controlled lighting, throughout a video sequence and precisely recovers the face position and orientation. We also propose animation techniques to embed realistic expressions in our 3D clones. Such face models permit automatic construction of appearance models.
展开▼