Intelligent feedback methods for multimedia databases
When doing content-based (as opposed to search entirely relying on annotation) search for images, there is a semantic gap between the user's query formulation capacities and the target of the search. We try to bridge this gap using techniques such as Bayesian Learning or Data Mining.
Software engineering for open multimedia systems
Having a database concept which allows seamless integration of both content based image retrieval (CBIR) and text information retrieval (IR), we are very interested in creating an architecture which permits us to make our system grow in every desired way with as little effort as possible. One step in this direction is MRML, an extensible language designed for the communication between the Multimedia Database and its interface. The goals are outlined in our draft technical report MRML: Towards an extensible standard for multimedia querying and benchmarking - Draft Proposal (it's a "living document", presently).