基于QR分解与2DLDA的单样本人脸识别  

Single Sample Face Recognition Based on QR Decomposition and 2DLDA

作　　者：覃磊[1,2] 李德华[1] 周康[2] 

机构地区：[1]华中科技大学自动化学院,湖北武汉430074 [2]武汉轻工大学数学与计算机学院,湖北武汉430023

出　　处：《微电子学与计算机》2015年第2期65-68,共4页Microelectronics & Computer

基　　金：国家自然科学基金项目(61179032;61072143);湖北省自然科学基金项目(2011CDB229);湖北省教育厅科技计划项目(Q20131703)

摘　　要：提出了一种新的基于矩阵的QR分解与2DLDA的单样本人脸识别算法(QR decomposition+2DLDA).利用矩阵的QR分解,将单样本人脸图像进行QR分解后提取有效的部分信息构成虚拟图像,结合原训练图像生成新的训练样本集,应用2DLDA进行特征提取和识别.在ORL人脸数据库上对算法进行了实验,实验结果表明此算法的识别效果不仅优于PCA、SPCA、(PC)2 A、E(PC)2 A算法,而且对于光照、表情等因素具有良好的鲁棒性.This paper proposes a new single sample face recognition algorithm based on matrix QR decomposition and2 DLDA.The algorithm uses QR decomposition of a matrix on the single sample,extracts the effective information to constitute a virtual image,combines the original training images and generates a new training sample set,then the application of 2DLDA algorithm for feature extraction and recognition.The algorithm is tested on ORL face database,experimental results show that this algorithm is not only superior to the traditional algorithms on recognition rate,such as PCA,SPCA,(PC)2 A,E(PC)2 A algorithms,but also has good robustness for factors such as illumination,expression etc.

关 键 词：虚拟图像 单样本 二维线性判别分析 QR分解 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]