摘要
针对图像中人脸部分比非人脸部分小得多的情况,设计了分类器专门用于人脸的粗定位。该分类器算法的主要步骤是,先对图像进行预处理,再使用OAC滤波器滤波,最后使用阈值分割法将图像分割为人脸区域、可能人脸区域和非人脸区域3部分。为人脸细定位算法减少了计算量,增加了准确率。对CMU数据库中的图像进行测试,非人脸区域的排除率能达到99%,误检率仅有1.3%。实验结果证明了该算法的有效性和可靠性。
Face detection is the first step of face recognition. This paper discusses the classifier based on the observation that presence probability of "face" objects in a scene is substantially smaller compared with that of "non-face" objects. The main processes of the algorithm are: to preprocess the image first and then filter in Fourier transform domain by OAC and finally segment the image into "face" area, "may-be face" area and "none-face" area. The algorithm reduces the computation of the second level classifier and increases the detection rate, The testing of CMU face database show that the algorithm successfully segments the image with a "non-face" rejecting rate of 99% and a false alarm rate of 1.3 %. The experiment results confirm the algorithms' effectiveness and reliability.
出处
《电子科技》
2006年第4期16-19,共4页
Electronic Science and Technology
关键词
人脸定位
OAC滤波器
模板训练
局部直方图均衡
图像分剖
Face detection
OAC filter
training template
local histogram equalization
image segmentation