THE RETRIEVAL OF IMAGES FROM IMAGE DATABASES - TRADEMARKS

This paper focuses on the problem of information retrieval from databa ses containing images rather than text. We propose an error-tolerant a lternative to menus and keywords-the feature-matching approach-in whic h users describe what they want to retrieve in response to a set of qu eries. The system matches the user's description with descriptions of images already in the database. Database images are then presented to the user in order of similarity to the user's description. The present paper serves four purposes: application of our feature-matching appro ach to a new image domain (trademarks); systematization of the process for developing these systems (articulation of five stages in the proc ess of system development); specification of criteria for selecting fe atures to maximize system performance; and introduction of concepts of power of discrimination and error tolerance to show how measures of t hese two factors can be used for evaluating system performance and opt imizing system development. Evaluation of the system (including experi ments on a small pilot database of trademarks and simulations of large databases) show the proposed set of 150 features would, without modif ication, be capable of handling expansion of the database to over 50 0 00 trademarks while still retrieving the target within the first 10 it ems on average. Analysis suggested several changes that should further improve the feature set.