I. Introduction
With ever more popularity of video web-publishing, much video content is being mirrored, reformatted, modified and republished. Such redundancy creates problems for multimedia search engines in that the search results become cluttered with a large number of similar versions of the same content. This degrades the usability of the search engine as studies show that users seldom go beyond the first screen of search results [2]. Content providers are also interested in using search engines to identify similar versions of their video content for legal, contractual or other business related reasons. To detect similar video content on the web, we require an algorithm to be capable of measuring sequences of any length, be robust against temporal re-ordering, and extremely efficient to execute.