Stochastic relaxation, Gibbs distributions and the Bayesian restoration of images*

摘要: We make an analogy between images and statistical mechanics systems. Pixel gray levels the presence orientation of edges are viewed as states atoms or molecules in a lattice-like physical system. The assignment energy function system determines its Gibbs distribution. Because distribution, Markov random field (MRF) equivalence, this also MRF image model. is more convenient natural mechanism for embodying picture attributes than local characteristics MRF. For range degradation mechanisms, including blurring, non-linear deformations, multiplicative additive noise, posterior distribution with structure akin to By analogy, defines another (imaginary) Gradual temperature reduction isolates low-energy (‘annealing’), what same thing, most probable under Gib...