An image restoration can be often formulated as an energy minimization problem. When an energy function is expressed by using the hamiltonian of a classical spin system only with finite range interactions, the probabilistic model, which is described in the form of Gibbs distribution for the energy function, can be regarded as a Markov random field (MRF) model. Some approximate optimization algorithms for the energy minimization problem were proposed in the standpoint of statistical-mechanics. In this paper, the approximate optimization algorithms are summarized and are applied to the image restoration for natural image.
Approximate Optimization Algorithms in Markov Random Field Model Based on Statistical-Mechanical Techniques
50_irai02.pdf
(518.15KB)
