Design And Real-Time Rendering Of The 3D Individual Face Model

3D modeling is currently one of the most important issues in computer vision and computer graphics. Human faces are so common and reachable that many new algorithms are first tested on the face data. Also, due to the complexity and flexibility of human faces, it has many challenging open problems. Among them, the personal 3D face modeling can be widely used in many virtual environments, such as the computer game, the virtual compere, the face animation, the film and advertisement etc, so many researchers have taken on the challenge of realistic 3D face modeling. At present there are many good methods of the neutral standard modeling, but they are difficult used in the personal face. The model built by the general modeling software is very realistic and lifelike, but it has complex structure and excessive data, which causes the problems that it can maintenance the quality of visual image, but has little feasibility of real-time rendering in the virtual scene roaming. These disadvantages cause the tremendous influence to the modeling, the memory, and the expense and quality of the rendering. Therefore, more and more people pay attention to this research area.Our work concentrates on the personal 3D face modeling and real-time rending in virtual scene. The study route of this paper is building up the neuter human face model's data structure, extracting the face features, improving the matching algorithm of individual face and the real-time rendering in virtual scene roaming. First, designs a Feature-Based Human Facial Model-FHFM based on the head model data provide by the third party software. Then adopts a method of mouse choosing, projection from 2D screen to 3D model, feedback of 3D coordinates to extract the coordinates of the featured points. Secondly, uses 3D coordinates of featured points synthesized from frontal and profile facial images to modify the predefined neuter face model. In this process presents a algorithm of degressive deformation by levels to realize the matching from neuter person face to the personal face. Finally, proposes a mesh simplification method related to the viewpoint and based on the blend of geometry and the surface characteristics, which realizes the real-time continuous multi-resolution rendering of the individual face model in the virtual scene.