基于三次B样条小波和2DFFT-2DLDA的人脸识别  

Face recognition based on cubic B-spline wavelet and 2DFFT-2DLDA

作　　者：陈勇[1] 冯浩[2] 王宪保[1] 

机构地区：[1]浙江工业大学信息工程学院,杭州310032 [2]嘉兴学院,浙江嘉兴314001

出　　处：《计算机工程与应用》2010年第6期151-153,159,共4页Computer Engineering and Applications

基　　金：国家自然科学基金(No.60871093)~~

摘　　要：提出了一种基于三次B样条小波和2DFFT-2DLDA的人脸识别方法,用三次B样条小波对人脸图像进行多层分解,得到一幅低频子图和3幅边缘细节子图,选取其中两幅效果最好的子图进行二维傅里叶变换后将其连接形成一个特征向量,然后进行2DLDA处理产生最终的特征表达,最后使用最近邻法进行分类。在JAFFE和Yale人脸库中的实验表明算法具有比频谱脸算法和Gabor-2DLDA算法更高的识别率,同时具有很低的算法复杂度。Face recognition based on cubic B-spline wavelet and 2DFFT-2DLDA is proposed.Cubic B-spline wavelet is used to decompose the facial image at suitable levels to produce an approximation subband and three detail subbands at last level composition.Then two optimal subbands selected from four subbands are performed 2D Fourier transform and concatenated into a feature vector,and then the vector is processed by 2DLDA to produce the last feature vector,the last result will be produced by the nearest neighborhood classifier.Compared with spectraface and Gabor-2DLDA algorithm,the tested result in JAFFE and Yale database shows that the proposed algorithm has significant performance.

关 键 词：人脸识别 小波变换 二维线性鉴别分析 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]