Identifying Video Content Consistency by Vector Quantization

Many post-production videos such as movies and cartoons present well structured story-lines organized in separated visual scenes. Accurate grouping of shots into these logical segments could lead to semantic indexing of scenes for interactive multimedia retrieval and video summaries. In this paper we introduce a novel shot-based analysis approach which aims to cluster together shots with similar visual content. We demonstrate how the use of codebooks of visual codewords (generated by a vector quantization process) represents an effective method to identify clusters containing shots with similar long-term consistency of chromatic compositions. The clusters, obtained by a single-link clustering algorithm, allow the further use of the well-known scene transition graph framework for logical story unit detection and pattern investigation.