Efficient feature representation employing PCA and VQ in the transform domain for facial recognition
Authors:
- 1 August 2007
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE) in Midwest Symposium on Circuits and Systems (mwscas)
- p. 281-284
Abstract
In this paper a new fast facial recognition system employing the principal component analysis, in the transform domain, and in conjunction with vector quantization, TD2DPCA/VQ, is presented. A transform domain two dimensional principal component analysis algorithm (TD2DPCA) was recently reported which possesses high recognition accuracy and low storage and computational requirements. The TD2DPCA/VQ presented here, maintains the recognition accuracy of the TD2DPCA while considerably improving the storage and computational properties. Employing the TD2DPCA/VQ, the storage and computational requirements are reduced by a factor P, where P is the number of training images (poses) per individual, used in the training mode. Experimental results employing the ORL and Yale databases confirm these excellent properties, where it is shown that the storage requirements and the computational complexity, for P=5, are reduced by 80% compared to the, high-performance, TD2DPCA algorithm.Keywords
This publication has 5 references indexed in Scilit:
- Recognition of Noisy Facial Images Employing Transform - Domain Two-Dimensional Principal Component AnalysisMidwest Symposium on Circuits and Systems (mwscas), 2006
- Two-dimensional pca: a new approach to appearance-based face representation and recognitionIEEE Transactions on Pattern Analysis and Machine Intelligence, 2004
- Face recognition using eigenfaces2002
- Vector Quantization and Signal Compression1992
- Eigenfaces for RecognitionJournal of Cognitive Neuroscience, 1991