Active vision-based face authentication

The use of biometric data for automated identity verification, is one of th e major challenges in secure access control systems. In this paper, several issues related to the application of active vision t echniques for identity verification, using facial images, are discussed and a practical system (developed within an European research project), encomp assing the active vision paradigm, is described. The system, originally devised for banking applications, uses a pair of act ive tracking cameras to fixate the face of the subject and extract space-va riant images (namely "fixations") from the most relevant facial features. T hese features are automatically extracted with a two-level algorithm which uses a morphological filtering stage for a coarse localization, followed by an adaptive template matching. A simple matching algorithm, based on a space-variant representation of fac ial features, is applied for identity verification and compared with a tech nique based on the Principal Component Analysis. Several experiments on identity verification, performed on real images, are presented. (C) 2000 Elsevier Science B.V. All rights reserved.