1. SVD-Based Projection for Face Recognition Chou-Hao Hsu and Chaur-Chin Chen Department of Computer Science Institute of Information Systems & Applications National Tsing Hua University, Hsinchu,Taiwan 30013 E-mail: cchen@cs.nthu.edu.twcchen@cs.nthu.edu.tw

2. Training Face Images Let F i (j) be the ith face image of m by n from the jth subject, 1 ≦ i ≦ N j and 1 ≦ j ≦ K, N 1 +N 2 +….+N k =N be the training face images. Define the mean image S as

3. Singular Value Decomposition S=UDV t Do S=UDV t where U and V are orthogonal. Select r,c with r ≦ m, c ≦ n such that d 11 +d 22 +…+d hh ≧ 85% of trace(D), where h=min{r,c} Let U r =[u 1,u 2,...,u r ], V c =[v 1,v 2,…,v c ] Where U is an m by m orthogonal matrix V is an n by n orthogonal matrix

4. Convert a face image into features For each training image A k, we represent this A k as x k =(U r ) t AV c, an r by c feature image For each test image T, we represent T by y=(U r ) t TV c, an r by c feature image

5. Distance between training feature images and a test feature image Compute d(y,x k ) by Fröbenius norm The smaller Fröbenius norm, the closer Rank the norms in an ascending order Determine the recognition rates from ranks 1, 2, 3,...,8 and plot the curve

6. Part of 5*40 Training Face Images

7. Missed Face Images and Their Wrongly-Best Matched Images

8. A Comparison of Difference Projection Methods