Statistical mechanics of digital halftoning

Abstract

We consider the problem of digital halftoning from the view point of statistical mechanics. The digital halftoning is a sort of image processing, namely, representing each grayscale in terms of black and white binary dots. The digital halftoning is achieved by making use of the threshold mask, namely, for each pixel, the halftoned binary pixel is determined as black if the original grayscale pixel is greater than or equal to the mask value and is determined as white vice versa. To determine the optimal value of the mask on each pixel for a given original grayscale image, we first assume that the human-eyes might recognize the black and white binary halftoned image as the corresponding grayscale one by linear filters. The Hamiltonian is constructed as a distance between the original and the recognized images which is written in terms of the threshold mask. We are confirmed that the system described by the Hamiltonian is regarded as a kind of antiferromagnetic Ising model with quenched disorders. By searching the ground state of the Hamiltonian, we obtain the optimal threshold mask and the resulting halftoned binary dots simultaneously. From the power-spectrum analysis, we find that the binary dots image is physiologically plausible from the view point of human-eyes modulation properties. We also propose a theoretical framework to investigate statistical performance of inverse digital halftoning, that is, the inverse process of halftoning. From the Bayesian inference view point, we rigorously show that the Bayes-optimal inverse-halftoning is achieved on a specific condition which is very similar to the so-called Nishimori line in the research field of spin glasses.