Linear Opponent-Colors Model Optimized for Brightness Prediction

摘要

Formal multivariate optimization techniques were applied in an attempt to determine how well a linear, opponent-colors model of color vision could account for specific brightness-matching data. The data fitted were from a single experiment by Sanders and Wyszecki that matched an adjustable white light in brightness to each of a set of lights of 96 different colors and constant luminance. A generalized, linear, opponent-colors model was formulated, which included the models of Guth (and co-workers), Ingling (and co-workers), and Thornton as special cases. The model contained 10 parameters, including nine determining the spectral responses of the three opponent-level channels and one determining the rule for combining the outputs of the three channels to obtain an estimate of equivalent luminance (the luminance of an equally bright white light). Despite difficulties with the optimization procedure, a model was found that correlates better than 0.98 with the fitted data. The predictions of the model for various other color-vision functions were explored and compared with corresponding predictions of the Guth and Lodge model and the Thornton model.