Retrieval of trade mark images by shape feature-the ARTISAN project

Many different approaches have been taken, some focusing on highly specific topics such as improved pattern-matching algorithms, others taking a much broader view, and attempting to study the issues involved by developing and evaluating prototype systems. Work on image retrieval at Northumbria University falls mainly into this second category. Our long-term aim is to investigate the principles on which workable shape retrieval systems can be constructed. Our approach has been essentially pragmatic to develop a series of steadily more sophisticated prototypes capable of handling an increasingly broad range of image types and retrieval paradigms. Our first project, the development and evaluation of SAFARI (Shape Analysis For Automatic Retrieval of Images), a prototype shape retrieval system for two dimensional engineering drawings of simple machined parts, demonstrated that we could achieve excellent retrieval performance within a restricted class of images. The aim of our current ARTISAN (Automatic Retrieval of Trademark Images by Shape ANalysis) project is twofold: to develop and evaluate a prototype shape retrieval system for trade mark images consisting of abstract geometric designs, and to assess the feasibility of generalizing the image retrieval approach developed in our earlier SAFARI system to a wider range of image types. It is hoped that the prototype system developed will provide a basis from which the Patent Office Trade Marks Registry can develop an improved image retrieval system for their own use