Classification of Near-Duplicate Video Segments in Broadcast Video by Occurence Patterns

Recently, technology to support the access to broadcast video using near-duplicate video segments has gained attention. By using near-duplicate video segments, it become possible to support the access to broadcast video contents such as construction of a commercial video database, tracking related news stories,structure analysis of a TV program. This paper proposes a classification method of near-duplicate video segments in broadcast video according to their usage. The proposed method refers to occurrence patterns to analyze and classify the categories of near-duplicate video segments. By an experiment with broadcast video from 6 stations for a week, near-duplicate video segments were automatically classified by the proposed method. The method presented the capability to generally handle individual application technology using near-duplicate video segments.