VideoGraph: A Graphical Object-Based Model for Representing and Querying Video Data

Modeling video data poses a great challenge since they do not have as clear an underlying structure as traditional databases do. We propose a graphical object-based model, called VideoGraph, in this paper. This scheme has the following advantages: (1) In addition to semantics of video individual events, we capture their temporal relationships as well. (2) The inter-event relationships allow us to deduce implicit video information. (3) Uncertainty can also be handled by associating the video event with a temporal Boolean-like expression. This also allows us to exploit incomplete information. The above features make VideoGraph very exible in representing various metadata types extracted from diverse information sources. To facilitate video retrieval, we also introduce a formalism for the query language based on path expressions. Query processing involves only simple traversal of the video graphs.