We are currently witnessing an explosion of interest in multimedia technology. Consequently, pictorial and video databases will become central components of many future applications. Access to such databases will be facilitated by a query processing mechanism that retrieves pictures based on user queries. This project considers research on the content based retrieval from pictorial and video databases. The resaerch is on a query processor that retrieves videos and pictures employing the meta-data describing their contents. The salient features of the video retrieval system is an expressive query language, called Hierarchical Temporal Language (HTL) , and a similarity based retrieval mechanism that can be built on top of any existing similarity based picture retrieval system employing meta-data. HTL uses the classical temporal operators to specify temporal properties of videos, and it employs level modal operators to specify such properties at different levels in the video hierarchy. At the atomic level, the language allows specification of properties on the meta-data associated with a single video segment (such as a frame or shot etc.). Given a user query, the query processing system employs the meta-data associated with the video segments to retrieve the videos that closely match the user query. This retrieval uses a similarity based semantics for HTL queries.
The other project in databases is on cost models in digital libraries, and on active and mobile databases, similarity based retrieval from sequence databases.