An advanced direct searching technique applied on compressed video content repositories

The DIVAS system is a content based search engine for locating multimedia fragments in large video repositories mainly found across the Internet. DIVAS performs searching using only audiovisual characteristics, unlike other, metadata-based solutions. Metadata based search methods are used as complementary services to DIVAS, in order to help and refine the overall search procedure, instead of using it as the primary search method. Thus it achieves a totally automated indexing and search scheme. This paper is focused on techniques adopted by DIVAS for analyzing, indexing and searching video content directly on the compressed domain. Significant improvement on search time can be achieved when avoiding decompression of the content which is one of the most time consuming processes when dealing with video processing. The generic DIVAS architecture is also presented in this paper highlighting its modularity and providing to the researcher and the developer the ability to easily add new features and to improve the performance and the accuracy of results.