Horizontal and vertical 2DPCA based discriminant analysis for face verification using the FRGC version 2 database

Jian Yang, Chengjun Liu

Research output: Chapter in Book/Report/Conference proceedingConference contribution

4 Scopus citations

Abstract

This paper presents a horizontal and vertical 2D principal component analysis (2DPCA) based discriminant analysis (HVDA) method for face verification. The HVDA method, which derives features by applying 2DPCA horizontally and vertically on the image matrices (2D arrays), achieves high computational efficiency compared with the traditional PCA and/or LDA based methods that operate on high dimensional image vectors (ID arrays). The HVDA method further performs discriminant analysis to enhance the discriminating power of the horizontal and vertical 2DPCA features. Finally, the HVDA method takes advantage of the color information across two color spaces, namely, the YIQ and the YCbCr color spaces, to further improve its performance. Experiments using the Face Recognition Grand Challenge (FRGC) version 2 database, which contains 12,776 training images, 16,028 controlled target images, and 8,014 uncontrolled query images, show the effectiveness of the proposed method. In particular, the HVDA method achieves 78.24% face verification rate at 0.1% false accept rate on the most challenging FRGC experiment, i.e., the FRGC Experiment 4 (based on the ROC III curve).

Original languageEnglish (US)
Title of host publicationAdvances in Biometrics - International Conference, ICB 2007, Proceedings
PublisherSpringer Verlag
Pages838-847
Number of pages10
ISBN (Print)9783540745488
DOIs
StatePublished - 2007
Event2007 International Conference on Advances in Biometrics, ICB 2007 - Seoul, Korea, Republic of
Duration: Aug 27 2007Aug 29 2007

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume4642 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other2007 International Conference on Advances in Biometrics, ICB 2007
CountryKorea, Republic of
CitySeoul
Period8/27/078/29/07

All Science Journal Classification (ASJC) codes

  • Theoretical Computer Science
  • Computer Science(all)

Keywords

  • Biometric Experimentation Environment (BEE)
  • Biometrics
  • Color space
  • Face Recognition Grand Challenge (FRGC)
  • Face verification
  • Feature extraction
  • Fisher Linear Discriminant Analysis (FLD or LDA)
  • Principal Component Analysis (PCA)

Fingerprint Dive into the research topics of 'Horizontal and vertical 2DPCA based discriminant analysis for face verification using the FRGC version 2 database'. Together they form a unique fingerprint.

Cite this