This chapter presents techniques for constructing the video "table of contents" (ToC), video highlights, and video index as well as the ways to integrate them into a unified framework. It reviews video analysis, representation, summarization, and retrieval. Further, a unified framework for video summarization and retrieval is described, followed by algorithms and experimental results on real-world video clips. Video content can be accessed by using either a top-down approach or a bottom-up approach. The top-down approach-that is, video browsing-is useful when one needs to get an "essence" of the content. The bottom-up approach-that is, video retrieval-is useful when one knows exactly what is needed from the content. In video summarization, the "essence" a summary should capture depends on whether or not the content is scripted. Because scripted content, such as news, drama, and movies, is carefully structured as a sequence of semantic units, one can get its essence by enabling a traversal through representative items from these semantic units. Hence, ToC-based video browsing caters to summarization of scripted content.
ASJC Scopus subject areas
- Computer Science(all)