We present a novel system for content-based image retrieval on the World Wide Web. The system consists of a keyword search and relevance feedback which can be applied to the results obtained by the keyword search. Contributions include clustering of several MPEG-7 Features and the combination of these features for relevance feedback. In the presentation
we will explain the clustering and the relevance feedback we are using in our system, as well as show some experimental results of our search engine.