A feedback-control model of human vision

In order for the human retina to achieve its very wide operating range (one billion to one in light intensity), high accuracy of spectral discrimination (the eye can distinguish among at least 10 million different shades of object color), constancy of object color, and uniform field of color perception, it appears necessary that the retinal receptors incorporate the following feedback-control processes: time-average feedback, spatial-average feedback, and automatic gain control. Time-average feedback would adapt each receptor to the time-average light it receives; spatial-average feedback would modify the signal from each receptor as a function of a weighted spatial-average of the receptor signals throughout the retina; and automatic gain control would keep the sensitivity of each receptor constant regardless of the adaptation conditions. This paper presents a model of a retinal receptor which incorporates these feedback control processes, and which appears to be consistent with physiological and psychological evidence.