Dichromatic color perception: a fast alternative for machine vision systems

摘要

This paper describes algorithms that discern chromaticity and luminosity in sensed images. The algorithms are inspired by the Guassian-like responsivity curves of biological photoreceptors. The algorithms work with pairs of tuned sensors, and color information extracted by the algorithms is dichromatic, not tri-chromatic as in human color perception. Over 50,000 simulations were performed. The simulations tested the effects of varying target images and several detector performance parameters including the means, variances, and absolute difference of the means of the photoreceptor responsivity. The principle conclusions drawn concern the algorithms' requirements for photoreceptor responsivity. After a brief introduction, a mathematical analysis of the algorithms is presented, simulation procedures are outlined, simulation results are analyzed, and conclusions are drawn.