2DLPP is a valid dimensions reduction method which directly extracts feature from image matrix and can detect the intrinsic manifold structure of data by preserving the local information of training data. We analyze the relation between 2DLPP and LPP. We demonstrate they are equivalent on some special conditions. Conventional 2DLPP is working in the row direction of images. We proposed an alternative 2DLPP which is working in the column direction of images. By simultaneously considering the row and column directions, we develop the two-directional 2DLPP, i.e. (2D)²LPP. The proposed method not only extracts feature with lower dimension than 2DLPP, but also take full advantage of row and column structure information of images. Experiment results on two standard face databases demonstrate the effectiveness of the proposed method.

Date Published: 13 January 2012
PDF: 7 pages
Proc. SPIE 8350, Fourth International Conference on Machine Vision (ICMV 2011): Computer Vision and Image Analysis; Pattern Recognition and Basic Technologies, 83501O (13 January 2012); doi: 10.1117/12.920530

Published in SPIE Proceedings Vol. 8350:
Fourth International Conference on Machine Vision (ICMV 2011): Computer Vision and Image Analysis; Pattern Recognition and Basic Technologies
Safaa S. Mahmoud; Zhu Zeng; Yuting Li, Editor(s)