Video index and search services based on content identification features

In this paper we examine the methodology, architectural issues and preliminary statistical results for identifying the presence and position of a given query clip within a massive collection of video content. This work is part of the European Union FP6 1ST Programme project DIVAS (Direct Video & Audio Content Search Engine). The concept is applicable to a number of use cases ranging from video clip search into large repositories, to DRM and content policing in the Internet.