The DICEMAN description schemes for still images and video sequences

Abstract

To address the problem of visual content description, two Description Schemes (DSs) developed within the context of a European ACTS project known as DICEMAN, are presented. The DSs, designed based on an analogy with well-known tools for document description, describe both the structure and semantics of still images and video sequences. The overall structure of both DSs including the various sub-DSs and descriptors (Ds) of which they are composed is described. In each case, the hierarchical sub-DS for describing structure can be constructed using automatic (or semi-automatic) image/video analysis tools. The hierarchical sub-DSs for describing the semantics, however, are constructed by a user. The integration of the two DSs into a video indexing application currently under development in DICEMAN is also briefly described.