This paper is a comparative study of three recently proposed algorithm
s for face recognition: eigenface, autoassociation and classification
neural nets, and elastic matching. After these algorithms were analyze
d under a common statistical decision framework, they were evaluated e
xperimentally on four individual data bases, each with a moderate subj
ect size, and a combined data base with more than a hundred different
subjects. Analysis and experimental results indicate that the eigenfac
e algorithm, which is essentially a minimum distance classifier works
well when lighting variation is small. Its performance deteriorates si
gnificantly as lighting variation increases. The elastic matching algo
rithm, on the other hand, is insensitive to lighting, face position, a
nd expression variations and therefore is more versatile. The performa
nce of the autoassociation and classification nets is upper bounded by
that of the eigenface but is more difficult to implement in practice.