Robust detail recognition and location technique for image retrieval in digital repositories

Large collections of image documents are today accessible via communication networks. Cost effective retrieval of the wanted images needs convenient browsing tools. In this contribution a technique for image selection on the basis of key details is illustrated. The technique is useful for instance for assisted queries to digital artwork collections, tele-sensing images, aerial photographs, astronomic databases, etc. The basic feature of the method is the capability of detecting and locating anyway rotated details with reasonable computational costs. In this work a criterion for proper choice of detail reference points is illustrated.