Face and Feature Finding for a Face Recognition System
A.W. Senior
In proceedings of Audio- and Video-based Biometric Person Authentication '99,
pp. 154-159. Washington D. C. USA, March 22-24, 1999.
This paper deals with the problem of nding facial features in images, a problem which arises in face recognition and in a number of other applications, especially in human-computer interaction, which derive information from human faces. This paper describes a system for nding faces in images and for nding facial features given the estimated face location. The techniques, based on Fisher's linear discriminant and distance from feature space, are presented, and results are presented on faces from the FERET database. The paper further describes how feature collocation statistics can be used to verify feature locations and estimate the locations of missing features.