Research On 3D Facial Texture Synthesis Based On Two Half-frontal Photos

The technology of network and multi-media has developed rapidly, which make people have increasing demand for human-computer interface. As a important part of 3d virtual person synthesis technology, with the carttoon workers'hard working for 30 years, 3d person synthesis technology has developed and improved rapidly. Beacasue 3D face synthesis technology has significance in theory and application, the technology has been used widely, and its achievements can be used in vitual conference, virtual host, Medical laboratories and research, film production, game entertainment, computer assistant instruction, and so on.This paper adopts binocular stereo vision principle, triangulation, affine transformation and kriging interpolation to study the generation of 3d face, the following results is reached:Firstly, build a real face sample library which is constiuted by 2d real face images and 3d real face models. Do preprocessing on 2d real face images and 3d real face models. Besides, real 3d face components are defined. Hence, the above three steps make 3d real face componet sample library be built.Secondly, two half-frontal photos are mapped to formal texture template through triangulation and affine transformation. Meanwhile, computer vision techs are used to implement the dimensional reconstruction based on the public regional feature points of the two half-frontal photos. Then the reconstruction model is vertical mapping to 2d space to form a 2d point set. Based on the 2d point set, an algorithm based on a face database produced by facegen is used to generate a similar three-dimensional model of the real face.