Structured patterns from random fields

The problem of modeling structured textures and image data with random field models is discussed. Random field models have recently been applied to a variety of problems such as image enhancement, segmentation, and coding. The models have been shown to successfully reproduce a variety of stochastic patterns, but have not been successful for capturing structure in patterns. Two possible solutions to the structured pattern problem are examined. The first solution is based on the structure present in the initial state of the random process, and the second is based on a relationship between internal random field interactions and coupling to a set of external constraints