Visual Routines " Oolo-0277/84r:$19.40 0 Elsevier Sequoia/printed in the Netherlands

This paper exlrmines the processing of visual information beyond the creation of the early representations. A fundamental requirement at this level is the capacity to establish visually abstract shape properties and spatial relations. This capacity plays a major role in object recognition, visually guided manipulation, and more abstract visual thinking. For the human visual system, the perception 6~ Jipatial properties and relations that are complex from a computational standpoint nevertheless often appears deceivingly immediate and effortless. The proficiency of the human system in analyzing spatial information far surpasses the capacities of current artificial systems. The study of the computations that underlie this competence may therefore lead to the development of new more eficient methods for the spatial analysis of visual information. The perception of abstract shape properties and spatial relations raises fundamental difficulties with major implications for the ojverall processing of visual informatioi. It will be argued that the computation of spatial relations divides the analysis of visual information into two main stages. The first is the bottomup creation of certain representations of the visible environment. The second stage involves the application of processes called ‘visual routines’ to the representations constructed in the first stage. These rou!ines can establish pro!lerties and relations that cannot be represented explicit/y in the initial representations. Visual routines are composed of sequences of elemental operations. Routines for different nroperties and relations share elemental’ operations. Using a fixed . set of basic operations, the visual syste:m can assemble different routines to extract an unbounded variety of shape properties and sp!atial relations. ‘This report describes research done at the Artificial Intelligence Laboratory of the Massachusetts Institute of Technology. Support for the laboratory’s artificial intelligence research is provided in part by the Advanced Research Projects Agency of the Department of Defense under Office of Naval Research contract NOOOI4-BOC-0505 and in paxt by National Science Foundation Grant 79-211IlOMCS. Reprint requests should be sent to Shimon Ullman B)epartment of Psychology and Artificial Intelligence Laboratory, M.I.T., Cambridge, MA