Leaflets Frontal Face Photo-based 3d Face Modeling Method For Automatically

Realistic 3-D face modeling has been a persistent challenge in computer graphics and computer vision field. Since the pioneer work of facial modeling was started by Frederic I.Parke in 1972, large amount of research work has been done on it and great progress has been achieved. The technique of Realistic 3D face modeling 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 great value in both theory and application.This paper discusses a 3-D face automatic modeling technique based on a single frontal face image. Only a single photograph with neutral expression and uniform illumination is needed in this method. It reconstructs 3-D model by using of the information of feature points detected automatically, then obtain the realistic 3-D face by texture mapping from the photograph. The research work mainly includes the following aspects:(1)Eye feature detecting: This paper advances a new approach for detecting the eyes' features: Based of face detecting, find out the area of eyes, then detect each feature point by using of the information of grayscale and gradient in this area. Because of determining the search area in advance, the location of the feature points detected is fairly accurate.(2) Lip feature detecting: Based on the research of skin-color and lip-color, A new method called WBRGC(Weighing Blue Color and Red Color Contrast Green Color)is presented in this paper, which improves the expression of the Red Exclusion. WBRGC discriminates skin-color and lip-color by taking advantage of the weighted average between the distribution difference of G,B component and G,R component. We acquire the best weight by experiment. The experiment result show that our approach's detecting ferformance is better than Red Exclusion's.(3) 3-D face modeling: By using of the key features detected, we make a series of modification on the general face mode (CANDIDE-3 model), then generate specific face model. Finally, synthesize the realistic 3D face by texture mapping.(4) We design a automatic face modeling system, which is implemented by VC++6.0, computer graphic library OpenGL and computer vision library OpenCV.