Based on physical laws of optics, this article analytically derives a
complete description of the set of all pictures that can be taken from
a given scene under varying lighting, where the camera, the scene, an
d the light sources are static but where each light source can vary ar
bitrarily in radiance. It will be shown that this picture set forms a
single convex region in picture space, where each picture is represent
ed as a high-dimensional vector. An optimum radiance invariant project
ion is derived which is invariant under a simultaneous change of the r
adiances of all light sources and which decreases the dimension of the
convex region by one. The theoretical predictions are confirmed exper
imentally for a scene illuminated with two light sources varying in in
tensity. A simple lighting invariant recognition algorithm is introduc
ed and tested in a face recognition experiment. A comparison between t
he new algorithm and a standard recognition algorithm is presented, wh
ich shows that lighting invariant recognition leads to considerably be
tter performance, even if the radiance of the light sources and the li
ghting directions change. (C) 1998 Academic Press.