Probabilistic Modeling to Inverse Halftoning via Multiple Dithered Images using Statistical Mechanical Informatics

Yohei Saika


Bayesian inference, Inverse halftoning, Monte Carlo simulation, Infinite-range model


On the basis of the Bayesian inference using the maximizer of the posterior marginal (MPM) estimate, we construct a method of inverse halftoning using multiple halftone images, so that the lower bound of the root mean square is inversely proportional to the number of the halftone images. Then, using the Monte Carlo simulation for a set of the snapshots of the Q-Ising model, we find that optimal performance is achieved under the Bayes-optimal condition, and that the Bayes-optimal solution reconstructs original images more accurately than the MAP estimate corresponding to a deterministic limit of the MPM estimate. These results are qualitatively confirmed by the analytical estimate via the infinite-range model. Further, we find that the present method accurately reconstructs realistic images by using the appropriate model prior.

Important Links:

Go Back