A framework for global integration of distributed visual information systems

With the advances in multimedia technology and the popularization of the Internet, it is now possible to access a large number of image and video repositories distributed throughout the world. The global integration of such databases enables users to access data in a transparent manner. In this dissertation, we present the design and implementation of a metaserver which supports content-based querying and intelligent selection of distributed visual databases over the Internet. The metaserver contains three main components: a metadatabase, a metasearch agent, and a query manager. To support intelligent site selection, the metadata for inclusion within the metadatabase is formulated on the basis of visual contents of the images housed at each remote visual database. The visual contents of the images in each database are summarized through image templates and statistical features characterizing the similarity distributions of the images. Given a visual query, the metasearch agent derives a list of potentially relevant databases by searching the metadatabase using a selection approach that uses query similarity to a template and the features of the databases associated with the template. The query manager transforms the information in the visual query for suitable matching of the metadata. Relevance feedback technique is used to incorporate user's response with the original query and produce a refined query for the subsequent database selections. The proposed framework has been implemented in a prototype system. Extensive experiments have been conducted to demonstrate that the designs presented in this dissertation collectively offer a viable and practical approach to distributed visual information retrieval and management.