基于分块离散余弦变换和Hausdorff距离的人脸识别  

Novel method for human face recognition based on block DCT and Hausdorff distance

作　　者：陈秀宏[1] 张伟伟[1] 薛寺中[1] 郝建东[2] 

机构地区：[1]江南大学数字媒体学院,江苏无锡214122 [2]解放军理工大学理学院,南京211101

出　　处：《计算机应用》2010年第12期86-89,共4页journal of Computer Applications

摘　　要：提出了基于分块离散余弦变换(DCT)和Hausdorff距离的人脸识别方法,利用分块DCT对人脸图像进行预处理,针对人脸图像存在扭曲变形的情况,提出了50%覆盖分块DCT的方法。提取DCT预处理之后的低频信息作为特征。最终利用Hausdorff距离和基于九点法的Hausdorff距离实现测试图像与模板图像的匹配。在ORL和Yale人脸库上反复实验表明,本方法不仅能够提高识别率,而且由于分块DCT快速高效的特点,能够显著缩短识别时间。This paper proposed a novel method for human face recognition based on block Discrete Cosine Transform（DCT） and Hausdorff distance.This method took pretreatment to facial images by using block DCT,and then proposed a method of 50% coverage on block DCT according to the distortation on facial images.After the block DCT pretreatment,the DC coefficient was taken out as recognition feature.In the end,the match of test images and templation was achieved by using Hausdorff distance and Hausdorff distance based on nine points formula.In order to evaluate the performance of this method,a series of experiments were performed on the ORL and Yale face image databases.The results demonstrate that this method can not only enhance the efficiency of recognition,but also shorten the recognition time significently,because of the high-efficiency of block DCT.

关 键 词：离散余弦变换 人脸识别 预处理 九点法 HAUSDORFF距离 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]