一种新颖的基于LDA的人脸识别方法.pdfVIP

第22 卷第5 期 红 外 与 毫 米 波 学报 Voi. 22 ，No. 5 2003 年l0 月 J. Infrared Miiiim. Waves October ，2003 A NOVEL FACE RECOGNITION METHOD BASED ON LINEAR DISCRIMINANT ANALYSIS ZHANG Yan-kun LIU Chong-oing （Institute of Image Processing and Pattern Recognition ，Shanghai Jiaotong University ，Shanghai 200030 ，China ） Abstract A novei face recognition approach based on the DCT and the LDA is proposed. First the DCT is used for dimen- sion reduction ，then LDA transform is performed on the iower space to extract feature. The ORL face database and the SJ- TU-IPPR face database are used to test our approach and the correct recognition rates of 97 . 5% and 92 . 6% are obtained respectiveiy，which shows that our approach is comparabie with other approaches. Key words face recognition ，iinear discriminant anaiysis ，discrete cosine transform ，principai component anaiysis. 一种新颖的基于LDA 的人脸识别方法 张燕昆 刘重庆 （上海交通大学图像处理与模式识别研究所，上海，200030 ） 摘要 提出一种基于离散余弦变换（DCT ）与LDA 相结合的人脸识别方法，首先利用DCT 将图像进行降维，然后在 低维空间中利用LDA 进行特征提取. 利用ORL 人脸数据库和我们上海交通大学图像处理与模式识别研究所的人 脸数据库进行测试，实验结果分别得到了97 . 5% 和92 . 6% 的正确识别率，表明它可以和其他方法相比较. 关键词 人脸识别，线性判别分析，离散余弦变换，主元分析. tures. But the first PCA step potentiaiiy ioses usefui Introduction discriminant information that is important for the foi- Face recognition has been of great interest in re- iowing LDA ［2 ］. Another drawback of the PCA is its cent years because its wide ra