Research On Content-based Video Retrievals

Multimedia Database has been the highlight among all kinds of study in the field of Database research, and yet there are still many technical obstacle to overcome. Content-based retrievals, or CBR, is totally a new concept and approach of the solution for retrieving data from vast amount of audio-visual (multimedia) content has been produced. As far as known, the most common way for searching data is text-based retrieval. Actually, however, it cannot meet the practical requirement, since there are so many data information such as brightness,contrast, consecutive frame, scenery and content binding in even a single video clip. In contrast, CBR can do better at this point.This thesis is primarily concerning building data model and data index in the theory part and feasibility of CBR in video retrieval in the technical realization part. Most of the research in this field are theory-oriented so far, so I mainly focused on the theoretical research.In the first chapter, the development of Multimedia Database and related technology were discussed firstly, then the research content and significance is provided. In the second chapter, my work is mainly concerning the fundamental research of how to build object-oriented data model & video data model, especially in the aspect of moldeling methods to construct an object-oriented video data model. The third part of this thesis discussed a lot about the approaches in building CBR index as well as the concepts & profound studys in (CBI) and (FBI). The aim of Feature-based index(FBI) is automatic index, and yet it cannot be realized because of current academic level in this field. So I primarily discussed CBI in detail in here. The fourth chapter illustrated the functionalities and their study approaches of three practical system,which were Content-basedvideo retrieval instances. In the fifth chapter, a.web-based sample content-based video retrieval system ,which was developed with the tools of ASP(Active Server Pages) programming technology and SQL server DBMS(Database Management System), was demonstrated. At the end of this thesis, the author made a brief summarization and pointed out the promising direction for further research of Content-based Video Retrieval.