Based On The Three-dimensional Reconstruction Of The Face Photo Of A Certain Angle Of Rotation

Realistic face modeling is one of the most difficult problems in computer graphics. With the development of virtual reality, character animation for films and advertising, Computer games, realistic face modeling attracts much attention of researchers. However, getting so much information is expensive in most situations. More importantly, it is difficult to collect enough information. So, this article will attempt to discuss 3D reconstruction of a non-front 2D face image.This article first introduced the application of 3D face model and the domestic and foreign research present situation, and analyzed and summarized the technology of 3D face modeling existed. And we define the face feature points, measuring points and the survey item based on the anthropometry, the anatomy etc.Second, estimated the degrees rotation of face picture, and authenticated by the CMU PIE face database. We establish the face database of the frontal and lateral face, to estimate the depth values of the feature points by using neural networks. Combining the feature points on specific face to make revision to the general model based on CANDIDE-3. Using the method of radial basis function (RBF) interpolation to make partial revision to get the non-feature points, estimated the depth values by using BP neural networks, generating the specific face model. And given one improvement of Loop segmentation method, to cause the segmentation of model conforms to the original shape, generating realistic face model by texture mapping.At last, the face database and face modeling system based on image are designed and implemented by Visual C++ and OpenGL.