Query processing for an MPEG-7 compliant video database

Abstract

Based on the recent advancements in multimedia, communication, and storage technologies, the amount of audio-visual content stored is increased dramatically. The need to organize and access the growing multimedia content led researchers to develop multimedia database management systems. However, each system has its own way of describing the multimedia content that disables interoperability among other systems. To overcome this problem and to be able to standardize the description of audio-visual content stored in those databases, MPEG-7 standard has been developed by MPEG (Moving Picture Experts Group). In this thesis, a query language and a query processor for an MPEG-7 compliant video database system is proposed. The query processor consists of three main modules: query parsing module, query execution module, and result fusion module. The query parsing module parses the XML based query and divides it into subqueries. Each sub-query is then executed with related query execution module and the final result is obtained by fusing the results of the sub-queries according to user defined weights. The prototype video database system BilVideo v2.0, which is formed as a result of this thesis work, supports spatio-temporal and low level feature queries that contain any weighted combination of keyword, temporal, spatial, trajectory, and low level visual feature (color, shape and texture) queries. Compatibility with MPEG-7, low-level visual query support, and weighted result fusion feature are the major factors that highly differentiate between BilVideo v2.0 and its predecessor, BilVideo.