Extraction of building product image from the Web

The HYPERCAT (HYPERmedia electronic CATalog) project proposes a digital organization of the technical information relative to the building products and materials. Its application on technical information search by images is one of context-based image search engines. However, a manual construction of an image database for this application can be very costly. The image extracted from the French building product providers' Web sites can solve the problem of acquisition and indexing. The problematics of Web image extraction for this activity are how can we extract and index the pertinent images. Consequently, this question leads to twofold challenges: image extraction and image indexing. First, the extraction rules are applied to illustrate the image extraction process. Second, the image indexing process uses the image context and a thesaurus. © 2004 Wiley Periodicals, Inc.