| Face is a very important and complex communication channel which plays an important role in human daily life.The face reconstruction provides an effective way to express and deal with face information,so the 3D face modeling is a very important for that.Binocular stereo vision realize the reconstruction of 3D scene by using the same method which human being binocular perceive the depth information of scene.This article use principle of binocular stereo vision to complete the 3D reconstruction of the face.The steps of reconstruction of the face are as follows:(1)This paper provides a comprehensive overview of the camera calibration technology based on the calibration method of Zhang Zhengyou.(2)The constraint conditions in stereo matching and stereo matching algorithm are analyzed in this paper,and then the image pairs are corrected by binocular stereo vision geometrical structure.Based on the traditional normalized cross-correlation matching algorithm,a bidirectional normalized cross-correlation matching algorithm based on recursive increment is proposed.In order to eliminate the mismatch points further,this paper proposes a binocular stereoscopic vision model by analyzing the binocular stereo vision model.The distance parameter of the matching point in the binocular system can be obtained by this model,which can effectively discriminate the mismatching points.(3)In the process of 3D face reconstruction an initial disparity map is filled by using the methods of region growing and then use a sub pixel-level interpolation algorithm to smooth the disparity map,and we obtain a complete and smooth disparity map.The 3D disorganized point cloud information of the human face is obtained by the optimized disparity map.Finally we get 3D reconstruction of human face by using triangulation.In the last part of the,a 3D face reconstruction application of binocular stereoscopic vision is developed by using joint-programming of HALCON and MFC. |
PDF Full Text Request