Abstract
Embedding color stimuli in colored surrounds changes their appearance. Such contextual effects could arise from the modulation of responses by lateral interactions in a population code for color (Wachtler et al. 2003, doi:10.1016/S0896-6273(03)00035-7). The strength of contextual effects in hue perception varies systematically with hue angle (Klauke & Wachtler 2015, doi: 10.1167/15.13.17). We present a model that explains this hue dependence by a population code with non-uniform distribution of preferred hues. The tuning distribution of model neurons was chosen according to the density of preferred hues of color-selective neurons in primary visual cortex (Wachtler et al 2003). Contextual interactions were modeled additively by subtracting the response of cells encoding the surround from the responses of cells encoding the stimulus with the same tuning preferences. Hue was decoded as the center of gravity of model neuron activity in hue space. Applying the model to stimulus-surround configurations with different stimulus and surround hues, the model showed a similar variation of contextual hue shifts as measured in psychophysical experiments (Klauke & Wachtler 2015). The model results were only weakly dependent on tuning width, which indicates a robust code given the variation of color tuning properties in primary visual cortex (Wachtler et al 2003). Furthermore, the output layer of the model showed a hue representation that was perceptually more uniform than the hue representation of a cone-opponent color space. In conclusion, non-uniform distribution of hue preferences could provide a physiologically plausible explanation of the properties of contextual effects in color perception.