Bidirectional 2D-OFD for Face Recognition
Un-Dong Chang, Kwan-Dong Kim, Young-Gil Kim, Youngjun Song, Jae-Hyeong Ahn · 2007 International Conference on Convergence Information Technology (ICCIT 2007) · 2007
In this paper, we propose an algorithm for face recognition. The new feature representation technique is called bidirectional 2-dimensional orthogonalized fisher discriminant analysis (B2D-OFD). B2D-OFD directly extracts image scatter matrix from 2D image and uses LDA for recognition. And then it eliminates dependence among feature vectors by orthogonalizing. By simultaneously considering both the row and column direction, we greatly reduce the size of feature matrix and get similar recognition rate. As a result of ORL database, the average recognition rate of B2D- OFD is 96.1%, while the average recognition rate of (2D)2FLD is 95.3% and the average recognition rate of (2D)2PCA is 94.6%.