Construction and use of research tools for image processing

Abstract

Image processing now has a wide variety of applications and a large amount of algorithm development is required. Clearly, a convenient and easily used development system is a great advantage. Some preliminary work with an existing machine indicated that a carefully tailored interactive facility could provide such an environment. An image storage unit containing a novel, fast, method of accessing the window to be processed has been constructed. By delegating to the storage unit some of the tasks normally performed by image processing software a considerable increase in processing speed has been achieved. While the improvement is not sufficient for an industrial system, it does allow for the convenient investigation of algorithms of considerably greater complexity than has hitherto been found possible on a moderately priced machine. To make full use of the hardware and to provide a concise notation for the description of processing algorithms, a versatile computer language, PPL2, has been developed. PPL2 provides, in addition to an extensive range of operators, a very concise yet very efficient method of denoting image operations. A compiler for this language has been incorporated into a complete image processing system for fast interactive development and testing of programs. Use has been made of the system to investigate the possible application of the quadtree in image processing and also for the formation of the skeleton description of an object. In the latter application interest centered around the possible advantages of a 5 x 5 over a 3 x 3 pixel window. Awareness of the potential industrial applications of image processing has led to observations and comments on the hardware and software required for image processing. Conclusions are reached concerning the relative merits of parallel versus sequential algorithms and of various types of processors.