Revisiting the Vector Retrieval Model in Context of the MPEG-7 Semantic Description Scheme

Abstract

Due to the semantic gap multimedia metadata is currently the only way to allow multimedia information retrieval on a conceptual level. One prominent multimedia metadata standard is MPEG-7. It offers means to annotate multimedia content using concepts and semantic relations. While several practical solutions for retrieval of MPEG-7 based conceptual annotations exist, a basic theoretical model has not been discussed yet. The paper in hand presents a vector representation for the MPEG-7 semantic description scheme and shows that the vector representation is unique and invertible. The vector representation allows the application of the vector retrieval model and therefore common retrieval mechanisms like weighting and indexing techniques or clustering.