Developing an Object-Oriented Video Database System with Spatio-Temporal Reasoning Capabilities

Content-based retrieval of Flash™ movies: research issues, generic framework, and future directions

“…In the VideoQ system proposed by Chang et al [5], videos are retrieved based on the joint spatio-temporal attributes of video objects represented as motion trajectories. In the work of Hjelsvold et al [14] and Chan et al [3], specialized query languages augmented with spatial and temporal operators are proposed for video retrieval. However, the complexity of the dynamic effects supported by Flash goes beyond the capability of the current techniques on modeling spatial/temporal features.…”

“…Two action types require more explanations: (1) "trace" is the action of an object following the movement of the mouse cursor in the movie frame; (2) "navigate" is an action of a Web browser being launched and directed to a specified URL, and therefore it does not involve any object. Compared with the existing models for spatio-temporal features, such as spatial/temporal operators [3], and motion trajectory [5], the concept of event provides a more compact, semantics-flavored representation of these features because the predefined action types directly address the human perception on the dynamic effects of a movie. On the other hand, it is also powerful in terms of expressiveness, mostly because an event can have multiple actions.…”