This paper addresses the problem of how to efficiently and effectively retr
ieve images similar to a query from a trademark database purely on the basi
s of low-level feature analysis. It investigates the hypothesis that the lo
w-level image features used to index the trademark images can be correlated
with image contents by applying a relevance feedback mechanism that evalua
tes the feature distributions of the images the user has judged relevant, o
r not relevant and dynamically updates both the similarity measure and quer
y in order to better represent the user's particular information needs. Exp
erimental results on a database of 1100 trademarks are reported and comment
ed. (C) 2001 Pattern Recognition Society. Published by Elsevier Science Ltd
. All rights reserved.