Computer vision and multimedia information systems
Thierry Pun, Ruggero Milanese · Archive ouverte UNIGE (University of Geneva) · 1995
Computer vision offers a number of techniques that can be used in the context of pictorial information systems.To motivate this point of view, we first present the basic issues involved in computer vision and in multimedia information systems, and emphasize the differences between these domains.We then outline the contributions that computer vision can make to the development of efficient multimedia information systems, especially for handling queries by visual example (QVE).In the second part of the paper, we concentrate on the matters of archiving and retrieving images from large databases, in the case of QVE.An approach is proposed that relies on two concepts, namely relevance and focusof-attention.These mechanisms allow to locate and select the most pertinent information from images.They can be used at the image archival stage, to automatically determine which parts of the image should provide meaningful indices to be compiled into the image access tables.Relevance and focus-of-attention are also important at the retrieval stage, to efficiently select from a database the images that best match a given pictorial query.