基于DCT的改进零空间人脸识别算法

赵传强; 王汇源; 吴晓娟

doi:10.3724/SP.J.1146.2006.01959

基于DCT的改进零空间人脸识别算法

doi: 10.3724/SP.J.1146.2006.01959

赵传强^{①②;王汇源},
王汇源,
吴晓娟

基金项目:

国家自然科学基金(60675024)资助课题

计量
- 文章访问数: 3450
- HTML全文浏览量: 106
- PDF下载量: 1002
- 被引次数: 0
出版历程
- 收稿日期: 2006-12-13
- 修回日期: 2007-09-28
- 刊出日期: 2008-07-19

Face Recognition Using Improved Null Space Method Based on DCT

摘要

摘要: 线性判别分析(LDA)是一种较为普遍的线性特征提取方法,它的主要缺点是在应用时经常遇到小样本问题,同时其准则函数并不与识别率直接相关。该文提出一种基于DCT的改进零空间LDA方法,能够解决以上两个问题。首先,通过使用DCT代替像素聚类并重新定义类间散布矩阵,得到一种新的零空间法。然后将这种方法与F-LDA结合起来得到一种新的对人脸识别更有效的特征提取方法,实验证明这种方法能得到较好的识别率。
- 人脸识别; 线性判别分析; DCT; 零空间; F-LDA
Abstract: Linear Discriminant Analysis(LDA) is one of the most popular linear projection techniques for feature extraction. The major drawback of applying LDA is that it often encounters the Small Sample Size(SSS) problem. Besides, their optimization criteria is not directly related to the classification accuracy. In this paper, an improved null space LDA method based on DCT is proposed to solve both problems. First, by employing the DCT instead of the pixel grouping and redefining the within class scatter matrix, a new null space method is given. Then, combining this method with F-LDA an efficient new feature extraction algrithm is proposed for face recognition. Experimental results show that this method achieves better performance than existing ones.

HTML全文

参考文献(1)

Chellappa R, Wilson C L, and Sirohey S. Human andmachine recognition of faces: A survey[J]. Proc. IEEE, 199583(5): 705-740.[2]Fukunaga K. Introduction to Statistical Pattern Recognition.Boston: Academic Press, 2nd edition, 1990: 592.[3]边肇祺, 张长水, 张学工. 模式识别[M]. 北京：清华大学出版社, 2000: 83-87.Bian Z Q, Zhang C S, and Zhang X Z. Pattern Recognition[M]. Beijing: Tsinghua University Press., 2000: 83-87.[4]Belhumeur P N, Hespanha J P, and Kriegman D J. EigenfacesvsFisherfaces: Recognition using class specific linearprojection [J].. IEEE Trans. on Pattern Anal. Mach. Intell.1997, 19(7):711-720[5]Turk M and Pentland A. Face recognition using eigenfaces.Proc. IEEE Conf. on Computer Vision and PatternRecognition, Maui, 1991: 586-591.[6]Yu H and Yang J. A direct LDA algorithm for highdimentionaldata-with application to face recogntion. PatternRecogntion, 2001, 34(10): 2067-2070.[7]Huang R, Liu Q S, Lu H Q, and Ma S D. Solving the smallsample size problem of LDA. Proceedings of InternationalConference on Pattern recognition, Canada, Quebec, 2002:29-32.[8]Chen L F, Liao H M, Lin J C, Ko M T, and Yu G J. A newLDA-based recognition system which can solve the smallsample size problem[J].Pattern Recognition.2000, 33(10):1713-1726[9]Loog M, Duin R P W, and Haeb-UmBach R. Multiclasslinear dimension reduction by weighted pairwise fisher criteria[J].IEEE Trans. on Pattern Analysis and MachineInterlligence.2001, 23(7):762-766[10]Lotlikar R and Kothari R. Fractional-step dimensionalityreduction[J].IEEE Trans. on Pattern Anal. Mach. Intell.2000, 22(6):623-627

施引文献

资源附件(0)

访问统计