基于分块pca 与稀疏表示的人脸识别face - bowen publishing.pdfVIP

Mar.2013, Volume 3, Issue 1, PP. 17-21 Scientific Journal of Computer Science 基于分块PCA 与稀疏表示的人脸识别 孙丽花，李云峰 河南科技大学机电工程学院，河南 洛阳 471003 摘 要：针对于全局人脸图像主元分析提取特征算法的计算复杂度，与识别方法对环境的敏感问题，提出一种基于分块图像 PCA 与稀疏分类的人脸识别算法。首先，将一幅人脸图像分为大小相同的子图像块，运用主成分分析对分块得到的子图像块 处理，构造其特征向量；再将测试图像某一块图像的特征矩阵用所有训练子图像的特征矩阵线性表示，利用L1 最优化求解稀 疏解，并计算各类训练样本对该子图像的残差值，最后根据相似度投票法进行分类识别。在YALE 与ORL 人脸数据库上验证 了该算法的有效性。 关键字：图像分块；人脸识别；线性表示；L1 最优化 Face Recognition Based on Modular PCA and Sparse Representation 1 2 Lihua Sun * , Yunfeng Li School of Mechatronics Engineering, Henan University of Science and Technology, Luoyang, 471003 sunlh1985@126.com Abstract-Considering the computational complexity of global PCA, face recognition based on modular PCA and sparse representation was proposed. Firstly, the original image was divided into the same size of subimage, the subimage were disposed by PCA and their eigenvectors were structured the eigenvector of a subimage of test image was represented linearly by the eigenvector of all subimage of train image values were calculated by the L1 optimization. the residuals of the subimage of all kinds of training samples were also calculated final classification and recognition was based on similarity voting. Finally, the effectiveness of the algorithm was verified by YALE and ORL face database was classified by a similarity voting method. Experimental results of the YALE and ORL face database showed that the proposed method is feasible for face recognition. Keywords- Image Part; Face Recognition; Linear Representation; L1 Optimization 引言 随着信息技术的快速发展，越来越要求人们的虚拟身份与自己的真实身份一一对应。由于生物识别技术具有 唯一性、可靠性和稳定性等性能，近年来被作为一种新的身份验证方法来研究，而且