双目下点云的三维人脸重建  

3-D point cloud face reconstruction with binocular system

作　　者：隋巧燕 董洪伟[1] 刘蕾[1] 

机构地区：[1]江南大学物联网工程学院,江苏无锡214122

出　　处：《现代电子技术》2015年第4期102-105,共4页Modern Electronics Technique

基　　金：中央高校基本科研基本科研业务费专项资金资助(JUSRP211A35)

摘　　要：提出一种在双目视觉系统下进行人脸重建的方法。首先,通过双目系统拍摄人脸的左右图像;其次,用Grab-Cut的方法的把人脸图像分割出来从而降低立体匹配的搜索范围;然后,用区域匹配算法得到人脸的视差图,从而得到人脸的三维点云;最后对不同角度的人脸图像进行SIFT特征提取和匹配。将提取的SIFT特征点和匹配关系反射到三维点云数据,获取不同角度人脸的三维点云数据的特征点和匹配关系,完成对不同角度的人脸进行粗配准。A method of face reconstruction by parallel binocular vision system is put forward. Firstly,photos of right andleft sides of human face are taken by binocular system. Secondly,human face images are segmented with Grab Cut method toshorten stereo matching search range. Thirdly,the regional matching algorithm（NCC）is used to obtain face disparity map and 3-D face point clouds. Finally,SIFT features of face images taken in different angles are extracted and matched,and then the ex-tracted SIFT feature points and matching relation are reflected to 3-D point cloud data to get feature points and matching relationof point cloud data of face got in different angles,and complete the coarse rectification of human faces in different angles.

关 键 词：双目视觉 图像分割 人脸视差图 三维点云 SIFT算法 

分 类 号：TN919-34[电子电信—通信与信息系统]