基于新颜色空间YCgCr的复杂背景图像下的肤色区域检测  

Skin-color Region Detection in Complex Scene Images Based on Novel Color Space YCgCr

作　　者：王建国[1] 华继钊[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8期229-231,共3页Computer Science

基　　金：国家自然科学基金资助项目(编号:60072034)

摘　　要：肤色信息在基于彩色图像的手势识别、人脸检测与跟踪和基于内容的不良图像过滤等应用中,起着非常重要的作用。为了有效地检测图像中的肤色区域,采用了类似于YCbCr颜色空间的新颜色空间YCgCr。为了说明YCgCr颜色空间的优越性,给出了该颜色空间与YCbCr颜色空间和Karhunen-Loeve(K-L)变换颜色空间中多样实验操作的比较。实验结果表明,用同样肤色样本得到的肤色阈值对相同的测试图像集进行肤色检测时,YCgCr颜色空间具有很好的肤色区域检测效果,漏检率和误检率均低于其它两个颜色空间的漏检率和误检率,并且对于不同的光照条件有较好的鲁棒性。Skin-color information plays an important role in applications such as hand gesture recognition,face detection and tracking and content based malignant iamge filtering based on color images. To detect skin-color regions effectively in images, a novel color space YCgCr that is similar to color space YCbCr is applied. To describe the advantages of the color space YCgCr, the experiment results achieved with this novel color space are compared with those obtained in YCbCr color space and Karhunen-Loeve （K-L） transform color space, the results on the same testing image sets demonstrate that YCgCr color space has more better skin-color region detection than that of the YCbCr color space and K- L color space based on the skin-color thresholds obtained through the same skin-color samples, false dismissal rate and false detection rate are lower than those of the two other color spaces and can work well under changing illumination intensity.

关 键 词：颜色空间 KL变换 肤色检测 错误率 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]