基于新颜色空间YCgCr的人脸检测与定位  

Face Detection and Location Based on a Novel Color Space YCgCr

作　　者：王建国[1] 王江涛[1] 杨静宇[1] 

机构地区：[1]南京理工大学计算机科学与技术学院

出　　处：《计算机科学》2007年第9期221-223,共3页Computer Science

基　　金：国家自然科学基金(编号:60072034)

摘　　要：提出了一种基于肤色的复杂背景彩色图像中人脸检测的方法。首先在新颜色空间YCgCr上对输入图像进行肤色检测并应用预处理技术缩小人脸检测的搜索区域;然后在物体区域方向计算的基础上,提出了能够检测任意旋转角度人脸的方法;最后对人脸候选区域采用模板匹配方法检测人脸,去除非人脸区域,降低误检率。实验结果表明,提出的方法对不同尺寸、任意旋转的人脸有较好的检测效果。Presenting a method for face detection based on the skin color in color images with complex backgrounds. Firstly, skin color of the inputed images is detected in the YCgCr color space and preprocessing techniques are introduced to reduce the searching areas of the face detection. Secondly, proposing a method that can detect arbitrary rotated faces based on the computation of the direction of the object. Finally, using template-matching method in the lace candidate regions to detect laces, removing the non-face regions, reducing the false detection. The experiment results show that the proposed face detection method has more better detection results to different scales and arbitrary rotated faces.

关 键 词：人脸检测 人脸方向 YCGCR颜色空间 模板匹配 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]