This PhD dissertation deals with knowledge-based systems and access to multimedia information. The contribution of the dissertation lies in the following Íelds: scene analysis, fuzzy knowledge bases expansion, access to fuzzy knowledge-based information retrieval systems, description and recognition of composite objects. Regarding scene analysis, (ch. 2), an improvement of a scene analysis mechanism, for encodedmotion picture was proposed. Extensive experiments onmovies - relatedmaterial gave an optimal selection of parameters. Results related to description and recognition of composite objects are included in ch. 3. An ontological framework, based on a description logic and compositional extension, is proposed, as well as a method to propagate uncertainty, based on evidence theory. Ch. 4, refers to the expansion of fuzzy knowledge bases with algebraic rules, algebraic closure and elimination of small elements. Furthermore, theoretical results about the properties of the equivalence closure and preservation of the properties of fuzzy binary relation after elimination of small elements are given. Results regarding access to fuzzy knowledge - based information retrieval systems are included in ch. 5. A method of estimation of the context of a query and for context - based query expansion is presented. This dissertation was prepared with TEXof Donald Knuth, LATEXof Leslie Lamport and W of Yannis Charalambous.