Topological principal component analysis for face encoding and recognition

Authors
Pujol, A Vitria, J Lumbreras, F Villanueva, JJ
Citation
A. Pujol et al., Topological principal component analysis for face encoding and recognition, PATT REC L, 22(6-7), 2001, pp. 769-776
Citations number
17
Categorie Soggetti
AI Robotics and Automatic Control
Journal title
PATTERN RECOGNITION LETTERS
ISSN journal
01678655 → ACNP
Volume
22
Issue
6-7
Year of publication
2001
Pages
769 - 776
Database
ISI
SICI code
0167-8655(200105)22:6-7<769:TPCAFF>2.0.ZU;2-3
Abstract
Principal component analysis (PCA)-like methods make use of an estimation o f the covariances between sample variables. This estimation does not take i nto account their topological relationships. This paper proposes how to use these relationships in order to estimate the covariances in a more robust way. The new method topological principal component analysis (TPCA) is test ed using both face encoding and recognition experiments showing how the gen eralization capabilities of PCA are improved. (C) 2001 Elsevier Science B.V . All rights reserved.