首页 | 本学科首页   官方微博 | 高级检索  
     检索      

基于曲率和积分投影的人脸特征检测
引用本文:胡静俊,陈国斌,刘济林,沈晔湖.基于曲率和积分投影的人脸特征检测[J].江南大学学报(自然科学版),2006,5(6):631-634.
作者姓名:胡静俊  陈国斌  刘济林  沈晔湖
作者单位:浙江大学,信息科学与工程学院,浙江,杭州,310027
摘    要:针对以往基于图像的人脸特征检测对光照敏感的缺陷,提出了一种基于3D数据的人脸特征检测算法.首先对人脸3D数据计算其主曲率,设定阈值进行人脸特征的初步分割得到二值化的分割图,然后在二值化图上进行水平和垂直积分投影进一步确定人脸特征的位置,最后利用人脸特征的先验拓扑知识对检测到的人脸特征进行验证.实验结果表明,该算法能够有效地提取人脸特征点,具有一定的鲁棒性.

关 键 词:人脸特征检测  曲率  三维人脸  积分投影
文章编号:1671-7147(2006)06-0631-04
收稿时间:2005-06-21
修稿时间:2005-09-08

Facial Feature Detection Based on Curvature and Integral Projections
HU Jing-jun,CHEN Guo-bin,LIU Ji-lin,SHEN Ye-hu.Facial Feature Detection Based on Curvature and Integral Projections[J].Journal of Southern Yangtze University:Natural Science Edition,2006,5(6):631-634.
Authors:HU Jing-jun  CHEN Guo-bin  LIU Ji-lin  SHEN Ye-hu
Institution:College of Information Science and Engineering, Zhejiang University, Hangzhou 310027, China
Abstract:To compensate for the illumination limitation of facial features detection in image,the paper is concentrated on the 3D facial features extraction.Firstly,surface curvatures,which can represent a local surface shape,are used in the segmentation.Secondly,horizontal and vertical integral projections are performed on the segmented facial image to find feature locations.Finally,feature locations should also be validated using the geometric constraints and symmetry constraints.The experimental results on real range images of several faces are provided.
Keywords:facial feature detection  curvature  range data  integral projections
本文献已被 CNKI 维普 万方数据 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号