有限空间低照度低分辨率图像中人脸检测  

Face Detection in the Low-Luminous-Density and Low-Resolution Images Within Limited Room

作　　者：孙继平[1] 陈伟[1] 唐亮[1] 刘晓阳[1] 林海涛[1] 

机构地区：[1]中国矿业大学煤炭资源与安全开采国家重点实验室,北京100083

出　　处：《中国矿业大学学报》2008年第3期373-378,共6页Journal of China University of Mining & Technology

基　　金：高等学校博士学科点专项科研基金项目(20050290010);国家电子信息产业发展基金重点项目(信部运[2005]555号)

摘　　要：为在有限空间低照度低分辨率监控图像中检测人脸,采用了模板匹配算法进行研究.用直方图均衡化增强和空域高通滤波方法对原始监控图像进行了处理;根据脸部固定结构特点,构造了平均人脸模板,并用模板匹配的方法来检测图像中的人脸.结果表明,空域高通滤波使图像的整体灰度和对比度降低,正面平均脸模板在匹配检测中适应侧面人脸的角度变大,而直方图均衡化图像的处理结果则相反;在此图像监控系统中,直方图均衡增强化适用于图像的输出模块;而空域高通滤波图像中的人脸检测的整体正确率比原图像的提高了2倍.For detecting the faces in the low-luminous-density and low-resolution monitoring images, the template matching method is adopted. The original images were processed with the histogram equalization and spatial-domain high pass filtering method. An average-face template is constructed according to the feature of the fixed framework of faces, and the faces were detected with the template matching method in the images. The results show that the whole gray level and contrast degree of images were reduced by the spatial-domain high pass filter. The angle of the side face was enlarged in matching detection with the obverse average-face templates. However, a contrary result would be obtained by the histogram equalization the contrary. In this image monitoring system, the histogram equalization can be used in the image output module, and the correctness rate of face detection in the spatial-domain high pass filtered image increased by two times.

关 键 词：底照度低分辨率图像 直方图均衡化 空域高通滤波 平均人脸模板 模板匹配 

分 类 号：TD6[矿业工程—矿山机电]