The need to retrieve visual information from large image collections is sha
red by many application domains. This paper describes the main features of
the multimedia information retrieval engine of Quicklook(2). Quicklook(2) a
llows the user to query image and multimedia databases with the aid of samp
le images, or an impromptu sketch and/or textual descriptions, and progress
ively refine the system's response by indicating the relevance, or non-rele
vance of the retrieved items. The major innovation of the system is its rel
evance feedback mechanism that performs a statistical analysis of both the
image and textual feature distributions of the retrieved items the user has
judged relevant, or not relevant to identify what features the user has ta
ken into account (and to what extent) in formulating this judgement, and th
en weigh their influence in the overall evaluation of similarity, as well a
s in the formulation of a new, single query that better expresses the user'
s multimedia information needs. Another important contribution is the desig
n and integration with the relevance feedback mechanism of an indexing sche
me based on triangle inequality to improve retrieval efficiency. The perfor
mance of the system is illustrated with examples from various application d
omains and for different types of queries (target search as well as similar
ity search). (C) 2001 Academic Press.