Abstract
We present a multimedia information analysis framework for content-based browsing of video. Specifically, we develop algorithms for the automated extraction of video highlights in sports video that are based on audio, text, and image features. The extracted annotations are used to build applications for selective browsing of sports videos. Such summarization techniques enable content-based indexing of multimedia documents for efficient storage and retrieval. In addition, in the context of the newly emerging standard MPEG-7, these methods will enable applications that use MPEG-7 descriptions. As this standard provides only the syntax for representing such descriptions and not specific algorithms for extracting them, these algorithms are of great value for establishing MPEG-7 as an accepted standard. We provide experimental results for the proposed algorithms on several hours of sports programs that prove the feasibility of efficient video access techniques in a multimedia environment.
