Since the context in which colors are seen on the monitor is very different
from that in which they are seen when printed, color imaging demands a rel
iable color appearance model. We present a method for faithfully approximat
ing the color appearance model CIECAM97s by means of feed-forward neural ne
tworks trained with the error back-propagation algorithm. In particular, we
show that it is sufficient to train a single neural network to approximate
the combination of the forward and reverse CIECAM97s models independently
by the adopted whites and the type of medium (CRT and hard copy). (C) 2001
Elsevier Science B.V. All rights reserved.