Contrast restoration model for contrast matching of cosine gratings of various spatial frequencies and areas.

The purpose of the model presented in this paper is to explain the well known fact that perceived contrast becomes independent of optical low-pass and neural high-pass filtering as well as areal integration with increasing stimulus contrast. In the model we assume that perceived contrast is computed by two different parallel mechanisms. One of them integrates signal information across space to improve the signal-to-noise ratio and is affected by the optical low-pass and neural high-pass filtering. The other mechanism estimates external local contrast by using inverse filtering. These two factors are combined by a 'restoration' mechanism so that the first mechanism affects the perception of low contrast and the latter that of high contrast stimuli. The behaviour of the model was tested against experimental results obtained with normal human observers. At low contrast levels, contrast matching curves were similar in shape to the detection threshold curves both as a function of the spatial frequency and area of the grating. At high contrast levels, contrast matches became physically correct. The model described the experimental results accurately.