基于分块离散余弦变换变换和主成分分析法的人脸识别.doc

毕 业 设 计（论 文） 题目 基于分块离散余弦变换和 主成分分析法的人脸识别 摘 要 随着计算机和网络技术的快速发展，信息安全呈现出前所未有的重要性。人脸识别因其广阔的发展前景和重要的理论研究价值，受到国内外众多研究机构的广泛重视。其中分块DCT变换是最近几年以来比较常见的图像处理工具，在人脸识别领域得到广泛的发展。 本文在研究了分块散余弦变换（DCT）的特点的基础上研究了基于分块DCT变换和PCA的人脸识别算法，充分利用了分块DCT变换快速高效的特点以及PCA方法降维减少运算时间的优势，大大缩短识别时间。首先利用分块DCT对人脸图像进行预处理，将DCT预处理后的低频信息作为特征。最后通过使用K-L变换取特征值、特征向量以及PCA降维的方式来形成特征脸，识别部分采用欧氏距离最近邻法进行了人脸识别。在ORL及YALE人脸库上实验结果表明，本方法不仅能提高人脸识别的识别率，并且利用了分块DCT快速高效的特点，能够显著缩短识别时间。 关键词：离散余弦变换(DCT)；主成分分析法(PCA)；K-L变换；人脸识别；预处理 ABSTRACT With the rapid development of computer and network technology, information security shows unprecedented importance. Face recognition because of its wide development prospect and important theoretical research value, extensive attention by many research institutions at home and abroad. The block DCT transform is more common in recent years since the image processing tools, has been widely development in the field of face recognition. Based on the study of the basic block discrete cosine transform (DCT) on the characteristics of the study on DCT and PCA face recognition algorithm based on full use of the block of DCT fast and efficient features and PCA dimension reduction methods to reduce the computation advantage of time, greatly reducing recognition time. Firstly, block DCT face image preprocessing, the low-frequency DCT preprocessed information as a feature. Finally, by using the KL transform taken eigenvalues??, eigenvectors and the way to form the PCA dimensionality reduction features face recognition part of the face recognition using Euclidean distance nearest neighbor method. The experimental results on the ORL and YALE databases show that this method can not only improve the face recognition rate, and can significantly save the identification time. Keywords: discrete cosine transforms (DCT); principal component analysis (PCA); KL transform; recognition; preprocess 目 录 前 言 1 第1章 人脸识别相关理论 2 1.1 人脸识别研究的背景及意义 2 1.2 人脸识别的现状及难点 2 1.2.1 人脸识别的现状 2 1.2.2 人脸识别的难点及现况 3 1.2.3 人脸识别的应用前景 4 1.