Functional VLSI Design for Low-Level Image Processing ICs, Etienne Allart, Georges M. Quénot and Bertrand Zavidovique, 22nd Annual Asilomar Conference on Signals,Systems and Computers, volume 2, pages 758-761, Pacific Grove CA, USA, 4-6 November 1988.

We propose a VLSI circuit design method which is derived from a functional decomposition of low level image processing algorithms. In order to design highly powerful architectures, we have tested this functional approach quite extensively and special purpose circuits have already been developped. Finally, from a broad class of various image processing algorithms and operators, we have defined a computational unit called functional element. Many of them are to be associated in order to form a mesh-connected programmable data flow computer. This data flow computer has been defined in order to get a straight mapping of algorithms expressed with the functional programming style proposed by Backus. Thus, high computing power can be easily obtained for real time complex image processing applications. Using the equivalent data flow graphs of the algorithms, this computer can help to determine whether a given combination of precisely specified IC operators would have the requested behavior, so that special purpose VLSI could be easily designed. Several examples of implementation onto this computer are shown in order to point out the salient concepts and advantages of this approach.