Face recognition incorporating ancillary information

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

Due to vast variations of extrinsic and intrinsic imaging conditions, face recognition remained to be a challengingcomputer vision problem even today. This is particularly true when the passive imaging approach is considered for robust applications. To advance existing recognition systems for face, numerous techniques and methods have been proposed to overcome the almost inevitable performance degradation due to external factors such as pose, expression, occlusion, and illumination. In particular, the recent part-based method has provided noticeable room for verification performance improvement based on the localized features which have good tolerance to variation of external conditions. The part-based method, however, does not really stretch the performance without incorporation of global information from the holistic method. In view of the need to fuse the local information and the global information in an adaptive manner for reliable recognition, in this paper we investigate whether such external factors can be explicitly estimated and be used to boost the verification performance during fusion of the holistic and part-based methods. Our empirical evaluations show noticeable performance improvement adopting the proposed method.

Original languageEnglish
Article number312849
JournalEurasip Journal on Advances in Signal Processing
Volume2008
DOIs
Publication statusPublished - 2008 Jan 1

    Fingerprint

All Science Journal Classification (ASJC) codes

  • Signal Processing
  • Information Systems
  • Hardware and Architecture
  • Electrical and Electronic Engineering

Cite this