Sensitivity study of estimation methods of the two-dimensional autoregressive model

Abstract

In this paper we present an estimator of the parameters of an AR-2D model that is an extension of an estimator presented for autoregressive models in time series. It uses an auxiliary model (BIP-AR) that limits the propagation of noise in an AR process. In addition, we present an analysis of the behavior of these new estimator (BMM-2D) and others estimators for the case of AR-2D processes contaminated by Gaussian noise. We also show an application to the image processing obtaining favorable results for our estimator. Computational implementation is carried out by R statistical software.