摘要
为了充分利用图像的肤色信息与人脸轮廓信息,本文提出了一种基于肤色预处理和Haar-like强度特征的人脸检测方法.首先,将图像转换到YCbCr色彩空间上,用肤色的多高斯阈值分割方法将肤色候选区域标定出来;然后用扩展特征集训练得到的级联分类器对肤色候选区域满足一定比例的待检测图像片进行检测,不满足的直接排除.实验结果表明,该方法比传统AdaBoost方法检测速度更快,检测率更高.
To make use of skin color information and face contour information, a new face detection method based on skin color preprocessing and Haar-like intensity features is proposed in this paper. First, skin color candidate regions are marked according to the method of Muhi-Gaussian threshold segmentation after converting the images into the YCbCr color space. Then, images will be detected by the AdaBoost algorithm with the extended set of features and those windows will be excluded if they don＇t reach a certain percentage of skin color candidate regions. Experimental results show that the proposed method provides better detection accuracy and achieves faster detection speed compared with the traditional AdaBoost method.
出处
《苏州大学学报（自然科学版）》
CAS
2011年第3期30-34,共5页
Journal of Soochow University(Natural Science Edition)
