A contrast- and luminance-driven multiscale network model of brightness perception.

A neural network model of brightness perception is developed to account for a wide variety of data, including the classical phenomenon of Mach bands, low- and high-contrast missing fundamental, luminance staircases, and non-linear contrast effects associated with sinusoidal waveforms. The model builds upon previous work on filling-in models that produce brightness profiles through the interaction of boundary and feature signals. Boundary computations that are sensitive to luminance steps and to continuous luminance gradients are presented. A new interpretation of feature signals through the explicit representation of contrast-driven and luminance-driven information is provided and directly addresses the issue of brightness "anchoring". Computer simulations illustrate the model's competencies.