Gabor特征结合Metaface学习的扩展稀疏表示人脸识别  

Extended Sparse Representation for Face Recognition Based on Gabor Features and Metaface Learning

作　　者：詹曙[1,2] 王俊[1,2] 方琪[1,2] 张启祥[1,2] 

机构地区：[1]合肥工业大学计算机与信息学院 [2]合肥工业大学安全关键工业测控技术教育部工程研究中心,合肥230009

出　　处：《光电工程》2014年第12期46-52,共7页Opto-Electronic Engineering

基　　金：国家自然科学基金(61371156);安徽省科技攻关计划(1401B042019)资助项目

摘　　要：为了克服人脸识别中存在的遮挡等闭塞问题,本文提出了Gabor特征结合Metaface学习的扩展稀疏表示人脸识别算法(GMFL)。考虑到Gabor局部特征对光照、表情和姿态等变化的鲁棒性,该算法首先提取图像的Gabor特征集;然后对Gabor特征集进行Metaface字典学习得到具有更强稀疏表示能力的新字典,同时引入Gabor闭塞字典来编码表示图像中的闭塞部分,并与新字典联合构造一组过完备字典基;最后利用过完备字典基求解稀疏系数重构样本,根据样本与重构样本之间的残差最小原则对人脸图像进行分类识别。在AR人脸库和FERET数据库上的实验结果验证了本文算法的可行性和有效性。To overcome the problem of occlusion in face recognition, the method of extended sparse representation for face recognition based on Gabor Features and Metaface Learning （GMFL） is proposed. Considering the robustness of the Gabor feature to the variation of illumination, expressions and gestures, the method extracts Gabor features of images firstly, and then a new dictionary with stronger sparse representation power can be obtained from the Gabor feature sets by Metaface scheme. Meanwhile, the Gabor occlusion dictionary is employed to encode the occluded part of the image, and a set of over-complete dictionary bases are produced. Finally, the test image can be reconstructed by the over-complete dictionary bases, and the residual between the sample and the reconstructed sample is used for classification by minimizing residual. Experimental results demonstrate that the algorithm proposed is valid and robust on AR database and FERET database.

关 键 词：人脸识别 稀疏表示 字典学习 GABOR特征 闭塞字典 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]