Detection of chromoluminance patterns on chromoluminance pedestals II: model

Abstract

A model for chromoluminance pattern detection and pedestal effects is described. This model has five stages. The stimulus is first processed by the cone array and then by color-spatial linear operators. The outputs of the linear operators may be expressed as weighted sums of cone contrasts over space. There are three opposite sign pairs of linear spatial operators in the model. Their spectral tuning at each point in space is similar to the luminance, green/red and blue/yellow mechanisms in color opponent models, but their sensitivity to cone inputs varies as a function of space. The operators in each pair are the same except that the signs of the cone inputs in one are the opposite of those in the other. A non-linear response operator follows each linear operator. It receives two inputs, one excitatory and the other divisive inhibitory. The excitatory input is the half-wave rectified output of one of the linear operators. The inhibitory input is a non-linear sum of all linear operator outputs. The non-linear response operator raises the excitatory input to a power, and divides it by the inhibitory input plus a constant to produce the response. The detection variable is computed by combining the difference in response to target-plus-pedestal and pedestal alone across the three non-linear operators. The model accounts well for the large data set presented in the companion paper and is generally consistent with other results in the literature. The spectral sensitivities of the inferred chromoluminance pattern mechanisms are similar to those obtained with different methods. The data set is shown to be inconsistent with several other models.