6. Conclusion

We have presented a comprehensive web-based video retrieval system, viz., VideoMAP*, and presented in this chapter its central mechanisms which offer a number of unique features. Particularly, we have discussed a methodology for identifying and extracting semantic information from media content which serves towards bridging the semantic gap between media content and its interpretation. The VideoMAP* system provides a hybrid approach to video retrieval over the web, by combining query-based method with content-based retrieval (CBR) functions. In addition, it supports an intermediate semantic scheme (viz., Papillon) for representing both full text and key frame semantics. VideoMAP* also provides concurrency control for virtual editing of video data among different users who are assigned with priority levels, and can accommodate various kinds of profiles containing knowledge information and activity models. These profiles are useful for annotating abstracted information into the video data and for processing queries of the video data. We have also shown the user interface facilities of our system from the perspective of sample user interactions.

There are a number of issues for us to further work on. Among others, we are extending our prototype system to make it as MPEG standard compatible, by allowing input videos to be in either AVI or MPEG format. We have also started to examine the possible performance issues, such as replication technique for the Global VPDB to be distributed into mirror databases, so as to accommodate larger number of Video Query Clients. Finally, the issues of optimizing CAROL/ST with CBR queries, and the possibility of combining local query processing with global query processing in light of overall query optimization are interesting for investigation in the context of VideoMAP*.