Realistic 3D Facial Synthesis

With the quick development of multimedia and network, the higher quality humanlike human-computer interfaces are required. As one of the most important parts of virtual human synthesis technology, 3D human face synthesis technique has been greatly developed in the recent 30 years. It can be used widely in the area of virtual reality environment such as virtual presenter, net meeting, videophone, remote education, remote medical diagnosis, movie effects and games etc. Research on 3D face synthesis has the great value in both theory and application.This dissertation researched some key problems in human face synthesis, such as 3D face modeling, MPEG-4 based facial animation, facial animation retargeting, animation with subtle texture, animation driving etc. The following are the main achievements in this dissertation:1) In the field of 3D human face modeling, the major problem is how to construct 3D face models correctly, easily and quickly. In this dissertation, an efficient 2D-to-3D integrated face reconstruction approach is introduced to reconstruct a personalized 3D face model from only one single frontal face image. This approach is fully automatic and the speed is fast. It is about fifteen times faster than the similar methods. This technique can also be used for the face recognition to expand the training data. The accuracy of face recognition is significantly improved, while the algorithms of face recognition are the same.2) MPEG-4 based 3D facial animation framework has the advantages of currency and few data. This dissertation presents a method to construct FATs(facial animation tables) for special human face. The FATs can be used to drive the standard face model defined by this dissertation and the models, which have the same topology with the standard model.3) Animation retargeting becomes a hot research topic of computer animation in the recent several years. Using the FATs for the standard model, this dissertation presents an approach to construct the FATs for any topology 3D face models. The method solves the problem of animation data reusing for any face models. Furthermore, it is highly automatic and the speed is quite fast.4) Traditional facial animation methods usually use a facial image with neutral expression, which cannot simulate the subtle texture changes on the face. In this dissertation,...