Mixed Limit Theorems for Pattern Analysis

Limit theorems are derived for probability measures of random configurations over graphs which are used as prior distributions in pattern theory. For one-dimensional graphs, these limits can be viewed as distributions of certain stochastic processes, while in higher dimensions the limits will in some cases have to be interpreted as belonging to Schwartz distributions. Such limit distributions are easy to use in pattern analysis, and greatly reduce the computing effort required in comparison with stochastic relaxation methods.