Maximum likelihood estimation of noisy Gaussian Markov random fields

M. Coli1, L. Ippoliti1
1Department of Quantitative Method and Economic Theory, University G. D''Annunzio, Chieti, Italy

Tóm tắt

In this paper we consider the issues involved in signal extraction and parameter estimation of particular spatial and spatio temporal processes observed with additive Gaussian noise. Within spatial statistics, we discuss maximum likelihood estimation of noisy auto-Gaussian models. For large lattices the estimation method can be computationally demanding thus, we present a maximum likelihood estimator which can be computed in O(n/sup 2/) steps. Spatio temporal processes are of main interest and parameter estimation of the STARG+Noise model class is also considered. The statistical properties of the proposed maximum likelihood estimator are finally explored in a simulation study.

Từ khóa

#Maximum likelihood estimation #Gaussian noise #Markov random fields #Parameter estimation #Lattices #Boundary conditions #Signal processing #Additive noise #Image sequence analysis #Image texture analysis

Tài liệu tham khảo

10.1109/TIT.1972.1054786 rue, 0, Fast Sampling of Gaussian Markov Random Fileds with Applications 10.1007/BF02589121 10.2307/1268381 ripley, 1981, Spatial Statistics, 10.1002/0471725218 mardia, 1979, Multivariate Analysis 10.1109/18.119691 ippoliti, 1999, Kalman Filter on Generalised STARMA Model ippoliti, 2001, On-line spatio-temporal prediction by a state-space representation of the generalised space time autoregressive model, Metron, 59 coli, 1999, Generalised spatio temporal models for environmental data, Proceeding of International Congress on Modelling and Simulation 'MODSIM05, 3 10.1093/biomet/64.3.616 besag, 1974, Spatial interaction and the statistical analysis of lattice systems, J R S S Series B, 36, 192 golub, 1996, Matrix Computations 3th Eds 10.2307/2291512 di giacinto, 0, Space-time noisy observations smoothing, Proceedings CLADAG2001 Meeting of the Classification and Data Analysis Group of the Italian Statistical Society Palermo di giacinto, 1994, Su una generalizzazione dei modelli spazio temporali autoregressivi media mobile (starmag), Proceedings XXXVII Riunione SIS San Remo, 11, 35 cressie, 1993, Statistics for Spatial Data 10.1002/jae.3950080504 dryden, 2002, Adjusted maximum likelihood and pseudo-likelihood estimation for noisy gaussian markov randomfields, Journal of Computational and Graphical Statistics, 10.1198/106186002760180563