基于DCT和分块2D2PCA的人脸识别  

Face Recognition Algorithm Based on DCT and Modular 2D2PCA

作　　者：张秀琴[1] 陈立潮[1] 潘理虎[1,2] 谢斌红[1] 

机构地区：[1]太原科技大学,太原030024 [2]中国科学院地理科学与资源研究所,北京100101

出　　处：《太原科技大学学报》2014年第5期333-338,共6页Journal of Taiyuan University of Science and Technology

基　　金："十二五"山西科技重大专项项目(20121101001)

摘　　要：为了解决人脸识别算法双向二维主元分析(2D2PCA)表征的信息不全面,鲁棒性差、识别速率较慢的问题,提出了一种结合二维离散余弦变换(DCT)算法和改进的双向二维主成分分析算法(模块(2D)2PCA)的新的人脸图像识别算法,该算法首先利用二维离散余弦逆变换(DCT)对人脸图像进行压缩,利用二维离散余弦逆变换(IDCT)对图像进行重建,可以去除了人脸图像中的干扰冗余信息。然后通过改进的2D2PCA算法即分块2D2PCA提取重建人脸图像中的特征。最后,用最近邻法对人脸图像进行识别,并定义了人脸图像相似度的概念。本文对ORL人脸图像数据库进行了实验。实验表明,本文算法有效的增强了识别的鲁棒性,缩短了识别的时间。In order to solve the problems in face recognition algorithm of bidirectional two-dimensional principal component analysis（ 2D2PCA）,such as uncomprehensive information,poor robustness and slower identification rate,this paper proposes a new face recognition algorithm combining two-dimensional discrete cosine transform（ DCT） with the improved two-directional two-dimensional principal component analysis algorithm（ modules（ 2D）2PCA）. The algorithm firstly compresses the face image by using two-dimensional discrete cosine inverse transformation（ DCT）. Using two-dimensional discrete cosine inverse transformation（ IDCT） to reconstruct the image can remove the redundant information of the face image. Then the features of face image reconstruction can be extracted by the improved 2D2 PCA algorithm called module 2D2 PCA. Finally,face image is identified by using the nearest neighbor method and the concept of face image similarity is defined. In this paper,experiments of ORL face image database are conducted,which shows that the new algorithm enhances the robustness of recognition effectively and shortens the time of the recognition.

关 键 词：人脸识别 二维离散余弦变换(DCT) 双向二维主成分分析((2D)2PCA) 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]