Visual adaptation of scale and imprecision in a noisy world

Abstract

Pointlike quantum noise in an image can be defeated either by representing the image at a low gray-scale resolution or at a low spatial resolution. The first solution locates an image at an inherent imprecision, and the second locates the image at an inherent spatial scale. Two vision-based models combat noise by automatic and local spatial-scale adjustment. Making contrast steps proportional to the square root of intensity (the deVries-Rose law) uses the imprecision solution. A hybrid system such as human vision use both spatial solutions, and also carries out similar processing in the time domain.