The development of information technology provides the accessibility to enormous numbers of multimedia objects

simple and complex. Indexing of multimedia objects has become a crucial issue. The variety of multimedia data model

the low efficiency of content based searching

and the lack of proper query language all have made multimedia indexing a difficult subject. The multimedia indexing project recently carried by the University of Kent at Canterbury (UK) has made a breakthrough in the related areas. The virtual database developed in the project provides users a tool to create and maintain a private view of interesting multimedia objects. The virtual database is organized in a hierarchical structure. The multimedia object references it stored are associated with a set of attributes called `cues' which are used in the browsing agent for constructing query conditions. The browsing agent performs the searching process on the user's behalf. Users first tell the agent how to find and recognize interesting objects

then set it going and control the query evaluation. The query specified by the user includes a framework

which plays the role of navigator in the evaluation process

and a collection of embedded queries. The browsing agent evaluates the query by checking the object cues stored in the virtual database and by calling individual external evaluators.