人脸识别中的眼睛定位方法  

Eye Location in Face Recogniton

作　　者：孙晓玲[1] 侯德文[1] 储凡静[1] 

机构地区：[1]山东师范大学信息科学与工程学院,山东济南250014

出　　处：《计算机技术与发展》2008年第10期46-48,52,共4页Computer Technology and Development

基　　金：山东省自然科学基金资助项目(Y2006G20)

摘　　要：眼睛是人脸部最重要的器官,因此眼睛准确定位在人脸识别中具有特殊重要的意义。基于灰度图像,提出一种改进的眼睛精确定位的方法。首先对图像进行光照和噪声的预处理,再利用积分投影和山谷法对图像进行分析,进而找出一系列的候选眼睛点,眼睛点的验证采用人脸的几何分布特征和眼睛模板匹配相结合的方法来进行最终眼睛对的确定。基于orl和yale人脸库的实验结果表明该算法对光照、旋转、遮挡不敏感,具有较高的定位准确率。Eyes are the most important organs of face, so locating eyes exactly play an important part in face recognition. A new method of eye location is presented in this paper based on gray images. After eliminating the image noise,the images are analysed on integration projection and valley method for locating the candidate eyes. The eyes are verified by geometric character and template matching. Experiments conducted on the orl and yale face-database show the algorithm is insensitive to illumination, rotation and covering,and it achieves excellent performance in detection rate.

关 键 词：人眼定位 人脸识别 积分投影 模板匹配 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]