English

应用科学学报 ›› 2004, Vol. 22 ›› Issue (4): 487-492.

• 论文 • 上一篇    下一篇

一种基于肤色和模板梯度的人脸检测方法

何良华1,2, 邹采荣2, 赵力1,2   

  1. 1 东南大学学习科学研究中心 江苏南京 210096;
    2 东南大学无线电工程系 江苏南京 210096
  • 收稿日期:2003-08-21 修回日期:2003-12-01 出版日期:2004-12-31 发布日期:2004-12-31
  • 作者简介:何良华(1977-),男,安徽枞阳人,博士生;邹采荣(1963-),男,江苏昆山人,教授,博导.
  • 基金资助:
    教育部《面向21世纪教育振兴行动计划》资助项目

A Method Using Skin-Color and Template Gradients for Face Detection

HE Liang-hua1,2, ZOU Cai-rong2, ZHAO Li1,2   

  1. 1. Research Center of Learning Science, Southeast University, Nanjing 210096, China;
    2. Department of Radio Engineering, Southeast University, Nanjing 210096, China
  • Received:2003-08-21 Revised:2003-12-01 Online:2004-12-31 Published:2004-12-31

PDF

78

被引次数

24

摘要： 提出了一种基于肤色和模板梯度的人脸检测方法,该方法首先利用肤色信息找到人脸的大致位置,再利用主分量分析提取出公共脸并以此作为匹配模板,最终采用二阶梯度法作为判别准则.结果表明本算法能明显地提高检测速度,检测率比一般的模板法有较大地提高.

关键词: 人脸检测, 肤色, 主分量分析

Abstract: Face detection is a traditional subject. However with the development of computer science, it is attracting increasing attention. In this paper a combined method for face detection is proposed using skin-color detection and template gradient matching. First, skin-color regions which are considered as face candidate areas are detected using a color model. The principal component analysis is then carried out to produce a common face which is used as a matching face model. Finally, the second order gradient is employed as a decision criterion. Experimental results show that the proposed method can substantially increase detection speed and improve the detection rate.

Key words: face detection, principal component analysis, skin-color

中图分类号: