Maximum entropy restoration of multispectral images using a deterministic quantum field model

The analogy between a multispectral image and a 2-dimensional solid in a thermodynamic quasi-equilibrium state is used to design a maximum entropy image restoration method. The solid is described by a dynamic quantum field model. Closed form solutions for the Gibbs partition function, the associated local Markov transitional probability functions and the resulting entropy function in equilibrium and quasi-equilibrium states of the 2-dimensional solid are given. Using these functions the entropy of the whole solid is maximized by maximizing the entropy of each of the single particles in the solid. The model is simple even for multispectral images because the calculations take place in the 2-dimensional solid model instead of in the multispectral image. The performance of the new method is illustrated by the restoration of simulated piecewise constant images with known noise characteristics. Well defined physical models of thermodynamics and statistical quantum field theory can be used to design powerful and simple models for image restoration