A major research subject in image databases is to support efficient an
d effective access to images based on their visual content. In color i
mage databases, this requires to support image retrieval by both globa
l and local chromatic features. Image retrieval by color regions is co
mplicated by the fact that regions which are requested in the query sh
ould correspond to relevant colored regions in the retrieved images. I
n the following we present the PICASSO system, which supports image in
dexing and retrieval based on colors. The system exploits a pyramidal
representation of images. Multiple descriptions of image properties ar
e created at different levels of resolution thus allowing effective re
trieval through specific queries as well as imprecise ones. (C) 1998 P
attern Recognition Society. Published by Elsevier Science Ltd. All rig
hts reserved.