基于肤色分割的人脸检测算法研究

Research on Face Detection Algorithm Based on Complexional Segmentation

介绍了目前人脸检测领域检测速度最快的Boosted Cascade人脸检测算法。该算法在进行人脸检测时没有考虑到肤色因素,在具体识别过程中,有些可利用肤色信息很快排除的区域,在Boosted Cascade算法中却没有被排除掉。针对该算法的缺陷提出了一种改进算法,即利用Boosted Cascade人脸检测算法,检测出人脸的候选区域,通过人脸肤色模型进行验证,如果候选区域的像素符合人脸的肤色模型的程度到达某一数值,则接受该区域,即认为该区域是人脸;否则排除该区域。改进后的算法能够有效地提高检测的正确率,减小出现检测错误的几率,在不影响识别速度的情况下,提高了检测效率。

This paper presents the highest speed face detection algorithm Boosted Cascade face detection algorithm. Since the Boosted Cascade algorithm hasn＇t taken the complexion into consider, it sometimes makes mistakes that can be easily solved when adding the complexional information to the algorithm. In order to solve the problems, this paper presents a solution of improving Boosted Cascade algorithm, that is detecting the possible face areas with the Boosted Cascade algorithm, using the model of complexional segmentation to verify the result of recognition. If the degree that the pixels of possible face areas accord with the complexional mode exceeds a certain number, then accept them, or abandon them. The improved algorithm has high detection accuracy, and it can improve the detection efficiency without slowing down the speed.