Conventional PCA usually uses total scatter matrix as a generation matrix, and two dimension image matrices must be transformed into vectors. In this paper, the between-class matrix generated by original image and its eigenvectors were used to feature extracting. First we compressed the image in horizon direction using 2DPCA, then we compressed the feature matrix in vertical direction. Thus, the dimension of features is lesser and the speed of classification is faster. At the same time the category information is fully used and the recognition rate are improved.

Date Published: 28 October 2011
PDF: 5 pages
Proc. SPIE 8205, 2011 International Conference on Photonics, 3D-Imaging, and Visualization, 82050Y (28 October 2011); doi: 10.1117/12.906299

Published in SPIE Proceedings Vol. 8205:
2011 International Conference on Photonics, 3D-Imaging, and Visualization
Egui Zhu, Editor(s)