Abstract

A particular problem for archive users is to locate and access appropriate material from 100,000s of hours of archive, for promotions, documentaries, etc. Automatic video analysis tools can assist by capturing and indexing metadata (invisible keywords that describe the recorded material). The project will develop an architecture in which these tools will be 'plugged-in' to a framework that allows them to be used across many applications. The project will also develop two such tools, for on-screen text recongition, and scene-cut detection.