Compressed Video Segmentation

Segmentation of video into shots and scenes in the compressed domain allows rapid, real-time analysis of video content using standard hardware. This paper presents robust techniques for parsing MPEG-compressed video sequences into shots based on their physical structure and further into scenes based on their semantic structure by identifying changes in content and camera motion. The analysis is performed in the compressed domain using available macroblock and motion vector information, and if necessary, DCT information. Motion vector analysis yields a qualitative description of the camera motion and is used to subdivide shots into subshots. Key frames for the shots and scenes can be used for browsing, indexing, and retrieval.