Robust RA Estimators in AR-2D Models for Images
Silvia Ojeda, Medina Allende · 1998
Abstract. Usually it is adopted for processing and analysis of images SAR, the multiplicative model Z ( m, n) X ( m, n) Y( m, n) = , where Z are the observations, X is the backscatter process and Y the noise speckle. In Frost et al (1982) (mentioned by Sant’Anna (1995)) a model AR-2D is suggested for the process X. Because the noise speckle is not cleaned totally, the robust methods are the most appropriate alternative to estimate this model’s parameters. The purpose of this paper is to present a proposal of robust estimation in the AR-2D contaminated model: The RA estimators. They conform the bidimensional version of the estimators based on the residual covariances for the ARMA unidimensional models, introduced by Bustos and Yohai (1986).