A digital simulation of a model for the processing of visual images is derived from known aspects of the human visual system. The fundamental principle of computation suggested by a biological model is a transformation that distributes information contained in an input stimulus everywhere in a transform domain. Each sensory input contributes under the transformation to every output, and every transformed output contains unique information derived from every sensory input. Perception is accomplished by paying attention to analytically predictable regions of the transform domain. The model thus far is capable of accurately classifying hand-printed alphanumeric characters, hand-drawn geometric shapes, typewritten Russian characters of varying quality, and small, frequently occuring words. A capability has thus been demonstrated for perceptual generalization tasks, noncontextual association tasks, and the reinforcement of classification decisions. One application of this which is still in the formulative stages would be the processing of complex aerial reconnaissance images. (Author/JY)