Abstract: This paper puts forward a new detection method by setting up skin-color template which is determined by the facial feature position and color information to solve the problem that the color-based face detection can only be used for rough detection at face region. Through pre-processing the image with illumination compensation and testing out face candidate region of the image by skin-color model in the YCbCr space, it confirms the position of the face by searching and comparing with the template constructed in this paper. Experimental tests show that the method has a better adaptability to various illumination environments and complex background picture, and the detection precision is also improved.

Key words: face detection, YCbCr color space, skin-color model, face template

摘要： 针对目前基于色彩的人脸检测只能用于人脸区域的粗检这一不足，提出一种利用人脸的五官位置及色彩信息建立彩色人脸模板的算法。采用光照补偿对图像进行预处理，利用YCbCr空间中的肤色模型进行粗检，确定出人脸候选区域，利用建构好的模板进行搜索比对定位出人脸。实验结果表明该方法对不同光照环境和复杂背景的图片均有较好的适应性，检测精度也得到了提高。

关键词: 人脸检测, YCbCr空间, 肤色模型, 人脸模板