4. Video Search

Given these rich MPEG-7 descriptions, it is possible to search for video content at a number of different levels, including features, models, and semantics [35]. The shot boundary description defines the basic unit of matching and retrieval as a video shot or segment. Users may issue queries to the system in several ways: (1) feature-based - by selecting example key-frame images and video segments in which matches are found based on the MPEG-7 feature descriptions, (2) text-based - by issuing text query which is matched against the MPEG-7 textual annotations or speech transcriptions, (3) semantics-based - by issuing text queries or selecting from key-words that are part of MPEG-7 classification scheme, and (4) model-based - by selecting key-words. The distinction between model-based and semantics-based is that the confidence score for model-based search can be used for ranking and fusing results during the video searching process. An example user-interface for MPEG-7 video search engine is shown in Figure 26.6.

click to expand
Figure 26.6: The MPEG-7 video search engine allows searching based on features, models, and semantics of the video content.