Recognizing Faces with Minimum Information

V.B. Lasca and D.L. Borges (Brazil)

Keywords

FACE RECOGNITION, PCA, SIGNIFICANT PARTS, PATTERN RECOGNITION

Abstract

Face recognition has been an area receiving great research attention lately. Hardware for image understanding tasks has become quite affordable and either for forensic, security, or surveillance purposes many face recognition systems had been proposed [2]. Although the performance of such systems has increased steadily one fundamental problem has been barely touched in the literature: is it possible to rank the most signi´Čücant features for faces and perform recognition with them? In this paper we propose a novel approach that extracts information from three small regions on the face only: eyes, nose, and mouth; the regions are ranked based on psychophysics and more empirical data, and a metric is proposed relating the three regions. Experiments were run using a database of 26 individuals in 4 different conditions each, ending up with 104 pictures. The results are promising and possible extensions are considered based on this work.

Important Links:



Go Back