The MAVIS project was a programme of research to develop Multimedia Architectures for Video, Image and Sound. Within the architecture separate modules are responsible for all the processing associated with a particular media-based feature type and, as new feature types are introduced and associated matching techniques developed, appropriate new modules may be added to the architecture. For example, to make use of the added richness which digital video presents, modules are being developed which understand the temporal nature of the video and which can extract combined spatial and temporal features.
The aim of the MAVIS 2 project, funded by EPSRC, is to introduce multimedia thesaurus (MMT) and intelligent agent support for content-based retrieval and navigation. The earlier MAVIS 1 project was concerned with enhanced handling of images and digital video sequences in multimedia information systems. The project will extend the Microcosm architecture to support the MMT, in which representations of objects in different media are maintained together with their inter-relationships. Intelligent agents will be developed to classify and cluster information from the MMT, as well as additional knowledge extracted during information authoring and indexing, to seek out discriminators between object classes and also naturally-occurring groupings of media-based features and to accelerate media-based navigation. Multimedia content-based retrieval and navigation also demands new media viewers which incorporate facilities for processing and analysis. Such viewers are being investigated, in particular to allow rapid identification of image-based objects.