For many sample-based texture synthesis algorithms using markov random field model, the size of neighborhood determines the quality and speed of texture synthesis.
In this paper, the structure of optimal policies for periodic Markov decision model is discussed and the linear program algorithms to solve it is given.