Modeling Video Data for Content Based Queries: Extending the DISIMA Image Data Model

We present an efficient video data model that extends the DISIMA image data model by adding the video components and setting up links between image and video data. Many video data models have been proposed, most of which describe video data independently of image data and therefore fail to consider the relationship between videos and images. Our proposed model expresses the semantics of video data content by means of salient objects and relationships among them. Connections between video data and DISIMA images are made through key frames, which are extracted from each shot. Based on these connections, techniques used to query image data may be used to query video data. In addition, a set of new predicates has been defined to describe the spatio-temporal characteristics of salient objects in the video data. MOQL is used as a query language, with which we present example queries that can be posed on the proposed video data model..