Segmentation, Index and Summarization of Digital Video Content
Di Zhong · 2001
Segmentation, Index and Summarization Digital Video Content Zhong this thesis, we propose develop unique frameworks and methods for temporal and spatial video segmentation object based video representation, indexing and retrieval both syntactic and semantic level. First we demonstrate a robust real-time temporal scene detection system that combines color, edge and motion features both compressed and uncompressed domains. Contrary existing work, considered comprehensive issues practical situations, such gradual transition, lighting change motion. algorithms perform very well variant kinds videos, including sports, sitcom, news, cartoon, movie and home videos. Then present automatic region segmentation system content-based video search. The system segments tracks consistent regions through each video shot, and then computes visual features extracted regions build visual libraries that support region level search. web-based video query system that has more than 3,000 video shots has been built. The query system allows users spatial-temporal search video shots drawing regions specifying features. the first video search engine that supports automatic extraction object-level motion-based search. Semantic object segmentation and tracking then studied to produce high-level object representation and description. introduce integrated scheme semantic object segmentation and content-based object search. AMOS, a unique video object segmentation system that combines low-level automatic region segmentation with user inputs is developed. object query model developed effectively combine local region-level features and spatial-temporal structures. This system very useful MPEG-4 MPEG-7 applications. end we present a real-time framework build semantic-level structure and event index live s...