Automatic Visualization of Story Clusters in TV Series Summary

摘要

This paper describes a visualization method for showing clusters of video stories for the purpose of summarizing an episode of a TV series. Key frames from the video story segments are automatically extracted and clustered based on their visual similarity. Important keywords are then extracted from video subtitles to describe the semantic content of each story cluster in the form of tag clouds. The evaluation of the automatic processing has shown promising results, as the generated summaries are accurate and descriptive.