230 The Content-Based Image Retrieval as an Assistance Tool to the Architectural Design Sabrina Kacher, Jean Claude Bignon, Gilles Halin and Gerald Duffing CRAI (Research Center in Architecture and Engineering) UMR MAP, N°694 CNRS Nancy School of Architecture 2 rue Bastien Lepage Nancy, France. ABSTRACT The architectural design requires a research of ideas and a documentation to help the designer in its creation work. It is a domain where the use of pictures (drawing, photography,…) is essential, because the information transmitted by photographic images are often more easily to understand than the one transmitted by texts. The goal of this work is to show the help that can bring the research of pictures indexed by visuals criteria, as colour, texture and shape, in the architectural design domain. If we accept the principle that " an image is better than 10000 words ", we can make the hypothesis that an image research indexed by visual criteria can bring a supplementary help to the designer when he tries to resolve design problems. We tested a research tool resting on image indexation with graphic attributes. Two types of corpora have been used. The first one contains images illustrating building products and the second one shows buildings or parts of buildings, which illustrate the wood architecture domain. The objective of this experiment is to evaluate the relevance of this type of image indexing according to identified users needs. We try to determine which type of visual criteria is the most appropriate to help the designer in the various phases of the design process. 1 INTRODUCTION The starting point of these research works is the assistance of Architecture design. More precisely, we work on the assistance to the information retrieval relative to the construction systems and building products. The target users belongs to the architectural domain, such as architects, designers… We identify three types of user's needs, which can be answered by developing a system of information retrieval by images: 1. The watching brief need: The user searches for new concepts without any preconceived ideas of the precise nature of the construction system or building products. He only wants to be informed of new solutions available by visualising images. 2. The vague idea need: The user searches for ideas, but he doesn’t know precisely what he is looking for. He visualises images and then during his navigation he starts to progress in solving its design problem. 3 The precise idea but no formulating need: The user knows precisely what he searches for, but he doesn't know precisely how to formulate it. For example the