The Transformation From Cone to Channel Sensitivities

Ingling, Carl

Abstract

Current models of color vision transform the signals from 3 cones, R,G,B to one summing (V lambda) and two differencing (r-g and y-b) channels.
The aim of the proposed research is to increase the predictive power of opponent models by working out more precisely the important features of the opponent transformation. These transformations are the core of all current color vision theories. In particular, the proposed research focuses on three problems: 1) Opponent models of color vision are not congruent with the X,Y parallel channel concept of electrophysiology. Our recent analysis of the spatial and spectral properties of the r-g channel show that this channel functions as an achromatic channel at high spatial frequencies and a differencing channel at low frequencies. In fact, the r-g channel must function as the major achromatic pathway for primate foveal vision, inasmuch as some 90% of foveal cells are X-cells. The proposed research aims to incorporate this discovery into a revision of conventional opponent theory, and to continue the experiments which so far have supported the theoretical analysis of the r-g channel. 2) Considerable data shows that the transformation from cone to channel sensitivities cannot be linear for either opponent channel. Nonlinearities are difficult to characterize. A physiologically-based construct, the silent surround, or theta-operator, shows promise in explaining a severe nonlinearity in the B-cone contribution to the r-g channel. Work on the theta-operator theory for the B-cone region of chromaticity space will be continued with the aim of accounting for the curvature of constant hue lines in this region. 3) The transformation equations must predict the curvature of lines of constant hue in the chromaticity chart (in fact, linear transformations require straight lines and hence are a priori incorrect). As part of the experimental program, we need to measure accurately lines of constant hue for certain key regions of the chromaticity chart. These data are crucial for testing theories of nonlinearity, and the existing data base is inadequate. Such data are necesssary for making further progress on equations defining the opponent transformations.