基于块分类器和SVM的人脸检测方法

针对彩色图像中人脸检测问题，文中提出一种由“块分类器 SVM分类器”组成的快速算法。它的主要特点是利用肤色点在YCrCb色度空间的分布概率模型，实现二值分割，然后利用面部区域双眼的典型非肤色特征，构造一个“块分类器”，从待检测图像中实现人脸候选窗的快速粗选，再用SVM分类器作进一步确认，最后对检测出来的窗口进行综合，从而实现人脸的精确定位。通过在相同测试集上的对比实验，在不影响正检率的情况下，该算法不但可以降低虚警率，而且极大地提高了检测速度。

Face Detection Method Based on Block Classifier and Support Vector Machine

In order to solve the problem of detection face of color images of real time,this paper proposes a novel and fast face detection approach to addressing this problem.The approach adopts a simple-to-complex strategy.First,use the skin-color probability model and the typical no skin-color character of the two eyes of the real face area to construct a block classifier,which can remove most face-dislike candidate from the complexity background of the image rapidly.Secondly,support vector machine-filter is applied to verify the face candidates abstracted by a block classifier.Last,merge the windows which havebeen verified in previous two steps as real face and then get the accuracy position for each face,thus can achieve the attainment of the face detection.In this way, all of the experimental results show that this algorithm is not only real time but also efficient.