基于多分辨率局部与全局特征融合的人脸识别  

Face recognition based on local and global feature fusion of multi-resolution

作　　者：阮小利 王顺芳[1] RUAN Xiao-li;WANG Shun-fang(School of Information Science and Engineering,Yunnan University,Kunming 650504,China)

机构地区：[1]云南大学信息学院,云南昆明650504

出　　处：《计算机工程与设计》2018年第9期2929-2933,共5页Computer Engineering and Design

基　　金：国家自然科学基金项目(11661081;11461079);云南省科技计划基金项目(2017FA032);云南省中青年学术和技术带头人后备人才培养计划基金项目

摘　　要：为提高低分辨率人脸图像的识别率,提出一种基于多分辨率局部与全局特征提取与融合的人脸识别方法。通过小波变换把低分辨率人脸图像分解为3种不同分辨率的人脸图像,分别采用快速PCA和Gabor小波对每一种分辨率下的人脸图像集进行全局特征和局部特征的提取。针对多分辨率局部与全局特征的融合提出不同的特征融合策略,对其得到的人脸特征选取Jackknife(刀切法)与k近邻分类器相结合的方法进行最终的分类和对比。实验结果表明,该方法对光照变化较大的低分辨率人脸数据集具有较高的识别性能。To improve the recognition rate of low resolution face images,a face recognition method based on multi-resolution local and global feature extraction and fusion was proposed.The low resolution human face image was decomposed into three different resolution face images by using wavelet transform.For face images of each resolution,the global features and the local features were extracted through the fast PCA and the Gabor wavelet respectively.Different fusion strategies were given to fuse multi-re-solution local features and global features.Jackknife test and K nearest neighbor classifier were combined to classify the obtained fusion features and to make the comparison.Experimental results show that the proposed method has better recognition perfor-mance for low resolution face data sets in the case of varying illumination.

关 键 词：多分辨率 人脸识别 融合 快速PCA GABOR小波 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]