摘要
本文分析了人脸的对称性和主成分分析法(PCA)、二维主成分分析法(2DPCA)的特性,证明了2DPCA协方差矩阵就是PCA协方差矩阵的主角线的平均值,同时表明2DPCA减少了对人脸识别有用的协方差信息。提出了一种基于人脸垂直对称性的变形2DPCA算法(S2DPCA),该算法最大程度地利用了协方差鉴别信息,用更少的系数表示一张人脸图像。通过在ORL的实验比较表明,该算法与PCA算法相比降低了计算复杂性,与2DPCA方法和PCA方法相比提高了人脸识别率,在识别率方面优于传统算法(PCA(Eigenfaces)、ICA、Kernel Eigenfaces),同时也压缩了人脸的存储空间。
In this paper the vertical symmetry of face,the characteristics of PCA and 2DPCA are discussed.And it is proved that the covariance matrix of 2DPCA is equivalent to the average of the main diagonal of the PCA covariance matrix,and eliminates the covariance information that can be useful for recognition.A reshaped 2DPCA algorithm based on the vertical symmetry of face（S2DPCA） is proposed which can make the most useful of the covariance discriminate information,represents a face with fewer coefficients.The experiments on the ORL face bases show it reduces the computational complexity compared with PCA,improve the recognition rate of face compared with PCA and 2DPCA,and is also superior to the traditional algorithms（ICA,eigenfaces and Keinel eigenfaces）,and shows a face image with fewer coefficients.
出处
《计算机工程与科学》
CSCD
北大核心
2011年第7期74-79,共6页
Computer Engineering & Science
基金
国家自然科学基金资助项目(60372049)
江西省科技计划青年基金项目(GJJ09412)
