Abstract
Ubiquitous networking infrastructures have been an effective means to facilitate the information delivery in the last century. Using web browsers, ones can retrieve numerous multimedia documents from remote web sites with no need of physical attendance. As fast-increasing volumes of multimedia contents brought on line, the refinement of the retrievals becomes of great significance to keep a principal focus of browsing. Traditional databases successfully employ alphanumeric attributes to manage data based on the equality. To augment such index capability into the multimedia content, many multi-dimensional index structures were proposed for multimedia data management. In this thesis, we will survey the recent progress of these structures, and discuss the underlying rationale behind the designs. Particularly, we examine the challenges to organize the high-dimensional multimedia objects. For quality measures, we develop several cost models to evaluate the recently proposed schemes, VA-file, IQ-tree, and UB-tree. Several empirical studies are performed to cross-examine the analytical findings. Finally, we conclude the investigation and discuss an alternative approach for possible performance improvement.