基于分块2DDCT和（2D）^2PCA的人脸识别  

Face recognition based on back 2D DCT and （2D）^2 PCA

作　　者：李文举[1,2] 尉秀芹[1] 高连军[1] 

机构地区：[1]辽宁师范大学计算机与信息技术学院,辽宁大连116081 [2]上海应用技术学院计算机科学与信息工程学院,上海201418

出　　处：《辽宁师范大学学报（自然科学版）》2013年第2期174-177,共4页Journal of Liaoning Normal University:Natural Science Edition

基　　金：辽宁省教育厅高等学校科研基金项目(L2010232)

摘　　要：人脸识别是生物特征识别技术中的重要研究领域,应用前景广阔.研究者们虽然提出了很多人脸识别算法,但其性能仍需进一步改进.为了提高现有人脸识别算法的识别准确率,提出了一种新的基于分块二维离散余弦变换(2DDCT)和双向二维主成分分析((2D)2PCA)的人脸识别算法.首先,将图像分块,利用2DDCT进行图像压缩,去除冗余信息,并通过逆2DDCT重建图像;其次,通过(2D)2PCA消除图像的行、列相关性,降低特征维数;最后,应用最近邻分类器进行人脸识别,在ORL人脸数据库中的实验证明了本算法的有效性.As an important field of Biometric Recognition, face recognition has wide application pros- pect. Many algorithms have been proposed, but their performance can be improved yet. To increase the accuracy rate of face recognition algorithms, a novel face recognition algorithm based on back two-dimensional discrete cosine transform （2D DCT） and two-directional two-dimensional PCA （（2D）^2PCA）is proposed in this paper. Firstly, a face image is segmented into block images, and the redundant information is dislodged by 2D DCT, and the images can be rebuilt via the inverse 2D DCT; Secondly, the row-column correlation of the images is dispelled and the feature dimensions are dwindled by （2D）^2PCA ; Finally, the face image is recognized by nearest neighbor classifier. The experiments on ORL database testified the availability of the proposed method in this paper.

关 键 词：人脸识别 二维离散余弦变换 双向二维主成分分析 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]