On the Correlation Structure of Nearest Neighbor Random Filled Models of Images

摘要

This paper discusses the correlation structure of some two-dimensional nearest neighbor random field models of images. The correlation structure of two nonequivalent random field models, the so-called simultaneous model and the conditional Markov model, are analyzed in continuous as well as discrete space. Analyses of two-dimensional moving average models and autoregressive and moving average models are also included. Based on the structure of the correlation function at lower lags an empirical test is suggested for the inference of image models. Examples from real textures are given. (Author)