Abstract
This study proposes how the visual cortex may process natural images under variable illumination conditions to generate surface lightness percepts. It is known that visual representations can adapt to a million-fold change in luminance. How such adaptation also “anchors” percepts of surface lightness to use the full dynamic range of neurons remains an unsolved problem. Such an anchoring of lightness helps to make an image look natural. Anchoring properties include articulation, insulation, configuration, and area effects (e.g., Gilchrist et al., 1999). A cortical model is developed that quantitatively simulates such psychophysical data, as well as psychophysical data about discounting the illumination gradient and the Cornsweet effect, among others. The model is also consistent with a range of anatomical and neurophysiological data about how the brain may use boundary representations to gate the filling-in of surface lightness via horizontal cortical interactions. The model filling-in mechanism runs a thousand times faster than mechanisms of previous biological filling-in models, and thereby helps to clarify how filling-in can occur at the speeds shown in perceptual experiments. The model can process natural images even under dim moonlight and dazzling sunlight. Application of the model to color domain illustrates that it is also able to process natural color images.