一种改进的Gabor算法及其在人脸识别中的应用  

The Application of an Improved Gabor Algorithm in Face Recognition

作　　者：薛晓利[1] 胡蓉[2] 朱金陵[1] 

机构地区：[1]西南交通大学电气工程学院,成都610031 [2]成都电子机械高等专科学校通信工程系,成都610071

出　　处：《成都电子机械高等专科学校学报》2011年第2期6-8,共3页Journal of Chengdu Electromechanical College

摘　　要：本文提出一种改进的Gabor小波变换方法——Sobel-Gabor人脸识别算法。人脸图片经过裁剪、对齐等预处理之后,先利用Sobel算子提取其梯度图,然后做Gabor小波变换得到表征人脸分类信息的特征矢量。利用卡方距离来计算矢量之间的不相似性,并采用最近邻分类器进行最终的分类判决。实验结果表明,同基于Gabor小波变换的人脸识别方法相比,Sobel-Gabor算法能显著提高识别率。This paper offers an improved face recognition algorithm,the wavelet transformation based on Sobel-Gabor.After the processes like crop and alignment,the face image gradients are first extracted by Sobel and then the Gabor transform is performed,getting the feature vector for the the final classification.The dissimilarity of the vectors is calculated by chi-square distance and the final classification decision is carried out with nearest neighbor classifier.The experiment shows that the algorithm is better in face recognition compared with the recognition based on Gabor transform.

关 键 词：边缘增强 小波变换 人脸识别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]