On the asymptotic variances of Gaussian Markov Random Field model hyperparameters in stochastic image modeling

This paper addrresses the problem of approximating the asymptotic variance of Gaussian Markov Random Field (GMRF) spatial dependency hyperparameters by deriving expressions for the observed Fisher information using both first and second derivatives of the pseudo-likelihood functions. The major contribution is that the proposed method allows hypothesis testing, interval estimation and quantitative analysis on the model parameters in several MRF applications, from image analysis to statistical pattern recognition. Finally, experiments using both Markov Chain Monte Carlo (MCMC) synthetic images and real image data provided good results.