基于图像分块和特征选择的单训练样本人脸识别  

SINGLE TRAINING SAMPLE FACE RECOGNITION BASED ON IMAGE BLOCKING AND FEATURE SELECTION

作　　者：李俊霞[1] 张书敏[1] 吴何胜 

机构地区：[1]河南农业职业学院电子信息工程系,河南郑州451450 [2]南京大学软件学院,江苏南京210093

出　　处：《计算机应用与软件》2015年第9期310-313,共4页Computer Applications and Software

基　　金：江苏省高新技术研究发展计划项目(BE2009142)

摘　　要：为了提高人脸识别的正确率,针对单样本人脸识别训练样本存在的缺陷,提出一种基于图像分块和特征选择的单样本人脸识别算法。首先将人脸图像划分成子块,并分别提取各子块的特征,连接成人脸图像特征向量,然后采用多流形判断分析算法选择对人脸识别结果贡献较大的特征。最后计算采用支持向量机对人脸进行识别,并采用Yale B和PIE人脸库对本文人脸算法的有效性和优越性进行仿真测试。仿真结果表明,相对于当前典型人脸识别算法,该算法提高了人脸识别正确率,获得了更加理想的人脸识别效果。In order to improve the accuracy rate of face recognition, we propose a single training sample face recognition algorithm which is based on image blocking and feature selection aimed at the defect of single face recognition training sample. Firstly, it divides the image into sub-blocks, and extracts the features of each sub-block separately to join them to eigenvector of face image, then it uses multi-manifold judgement and analysis algorithm to choose the features with greater contribution to face recognition result; The final calculation adopts SVM to recognise the face, and uses YaleB and PIE face library to carry out simulation test on the effectiveness and superiority of the proposed algorithm. Simulation result indicated that relative to current face recognition algorithm, the proposed algorithm improved the accuracy rate of face recognition and obtained more ideal face recognition effect.

关 键 词：训练样本 人脸识别 图像分块 特征提取 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]