Face Recognition

Abstract

Face recognition includes both identification and verification/ authentication. It involves matching the biometric signatures derived during enrollment against those derived for some unknown face probes. The biometric signatures consist of representations induced by the face space used. Matching requires access to distance functions to measure for similarity between the above signatures and a priori thresholds for making decisions regarding either verification or identification. Such decisions include acceptance or rejection for 1-1 verification, and (a) rejection or (b) detection and classification for 1-M identification (see Sect. 1.1 and Fig. 1.2). Verification checks if the biometric signature presented is genuine and belongs to the identity declared by the client. Identification usually corresponds to closed set face recognition and is implemented as iterative verification. It always finds a mate for identification, the most similar one, even if there is none. Rejection, an option available only during open set face recognition, is appropriate for the case when there is no enrolled mate for the unknown probe.