I am pursuing my research in Texture Interpolation. After digging deep into it, I'm stuck at a point. It is proven that textures can be modeled as Markov Random Fields and that each texture can be synthesized from a collection of neighborhoods. How exactly do we get these neighborhoods that can characterize a texture and if we have them, how can we synthesize the texture back ? And, how does MRF help in any of the above ? Can anyone please provide me with a explanation that helps a layman understand concepts above ?