基于肤色的复杂环境下人脸检测  

Face detection based on skin color with complex scenes

作　　者：王建国[1] 李利国[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：《计算机工程与设计》2007年第21期5153-5155,共3页Computer Engineering and Design

基　　金：国家自然科学基金项目(60072034)。

摘　　要：对由光源颜色变化引起的图像色彩偏差,进行了校正,并在YCbCr颜色空间建立了Cb-Cr色度查找表和亮度信息联合的肤色模型,应用预处理技术,去除部分非人脸区域,减少人脸检测的搜索空间,并采用模板匹配方法在人脸候选区域检测人脸。实验表明,该方法能够有效的从复杂环境的彩色图像中检测出左右旋转不超过45°的人脸,且不受人脸表情、尺度和数目的影响,且错误率较低。The image color bias produced is corrected by the illuminant color. A skin-color model ofCb-Cr chroma lookup table combined with the brightness is constructed. Pre-processing techniques are adopted to remove some non-face regions and reduce the searching re- gions for face location. Finally, template-matching method is used in the face candidate regions to detect face. Experiments demonstrate that the proposed method can detect faces of rotated angle in 45 degrees from the color images with complex scenes effectively, and is invariant to face emotion, scale and number, with lower error rates.

关 键 词：肤色模型 人脸检测 模板匹配 YCBCR颜色空间 复杂环境 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]