Intensity-dependent spread-a theory of human vision and a machine vision filter with interesting properties

Abstract

Summary form only given. Extensive psychophysical evidence indicates that the spatial and temporal filtering properties of the human visual system depend on local retinal image illuminance; as illuminance decreases, signals are integrated over larger areas and longer time. A model that reproduces the spatial effect is as follows. There are three layers, an array of photodetectors, a spreading network, and an array of output channels. The output of each photodetector spreads its signals in the network, in a way to be described, and the signal leaving each point in the output array is the sum of signals arriving at that point. If the point spread function were constant, this system would simply act as a spatial filter, convolving the spread function with the input.<>