基于肤色信息和主分量分析的人脸实时检测系统

Real-time Face Detection System Based on Skin Color Information and PCA

提出了一种基于肤色模型、人脸面部结构和模板匹配的人脸检测算法。该方法首先建立肤色模型来对原始图像进行肤色区域分割,再根据人脸的面部结构特征对分割区域进行过滤,最后用基于主元素分析的模板匹配算法对候选区域进行判断。实验结果表明该算法能够适应复杂背景以及多人脸的检测,而且检测速度快,能够达到实时检测的目的。

In this paper, one novel algorithm for face detection in color images by skin model, feature of face structure and principal component analysis is proposed. Firstly, skin color model is constructed by using a large database of labeled skin pixels in YcbCr space. Then the images are segmented to face candidates and background using the skin model, and the face candidates are validated using structure features of faces. Finally, model match method with principal component analysis is used to the face candidates. Experimental results with VC++ show that the proposed method can detect single or multiple faces from complex background in real time.