Processing Cones: A Computational Structure for Image Analysis.
Allen R. Hanson, Edward M. Riseman · 1981
Abstract : A layered hierarchical parallel array architecture for image analysis applications, referred to as a processing cone, is described and sample algorithms are presented. A fundamental characteristic of the structure is its hierarchical organization into two-dimensional arrays of decreasing resolution. In this architecture, a protypical function is defined on a local window of data and applied uniformly to all windows in a parallel manner. Three basic modes of processing are supported in the cone: reduction operations (upward processing), horizontal operations (processing at a single level) and projection operations (downward processing). Complex image analysis algorithms are specified as temporal sequences of function applications. The cone structure forms the basis of a sophisticated operating system for an image analysis environment. (Author)