The Study Of Complete Virtual Human Head Modeling And Facial Expression Synthesis

With the rapid development of virtual reality technology, computer graphics technology and computer vision technology,3D virtual human gradually develop into a new subject which has attracted more and more people engaged in related research. As a result of human facial expressions and emotions are very rich, the study of 3D virtual human head has become a hot spot in research. The research of 3D virtual human head not only has great research value, but also has broad application prospects. We can always see its shadow in the fields such as cosmetic medicine, e-commerce, teaching video and game development. In the research, people often establish the head model using 3D scanner or modeling software, but there always exists some problems with the established model using these methods. The most important are the amount of data is too large, not easy to control. Therefore, establishing a head model with relatively small amount of data, completely head structure and easy control has become the focus of the research in this thesis.With the increase of personnel and accumulation of related work, a number of different head models have emerged. These models are either has inadequate structures or the model is too complex and uneasy to control. To solve this problem, in this paper, we will establish a model with complete structure which can express the features of the head. Firstly, after simplifying the software established model, removing the mesh of the face, and combining the standard Candide-3 part together, we synthesized a general model. Then, the eyelid will be reconstructed using Bezier curve to simulate the edge inside and outside eyelids and upper eyelid sulcus. At the same time, the method of combining two different radius of sphere is used to establish the eye model, and with four corners of the patch to simulate the tooth, you can get a complete virtual human head model. After that, using the information of the feature points extracted from the inputted single frontal face image of facial model to adjust the facial feature points, non-teature points and hindbrain spoon points, we can get a specific head model which matches to the inputted image’s face, using different methods to texture mapping for different modules, you can get a real head model. Finally, using facial action units to adjust the vertex coordinates of the head model which related to the face and synthesized facial expression, a realistic facial expression animation is synthesized. Experimental results show that the established head model can give people a strong sense of reality, the model is easy to control, and the synthesized facial animation based on the model is comparably real.