Thesis Estimación robusta de parámetros en distribuciones con datos SAR
Loading...
Date
2003
Authors
Journal Title
Journal ISSN
Volume Title
Program
Magíster en Ciencias de la Ingeniería Informática
Departament
Campus
Campus Casa Central Valparaíso
Abstract
Muchas aplicaciones utilizan distribuciones asimétricas para ajustar datos reales y, frecuentemente, las situaciones prácticas requieren técnicas de inferencia paramétrica robusta. Esta tesis presenta la derivación de los M-estimadores con funciones de influencia asimétricas, motivados en una distribución basada en el modelo multiplicativo para imágenes contaminadas con ruido speckle. Esta ley puede ser altamente sesgada, y la estimación por máxima verosimilitud puede ser severamente obstaculizada por pequeños porcentajes de outliers. Los outliers aparecen principalmente debido a que las hipótesis de independencia e igual distribución de las observaciones rara vez son satisfechas en la práctica; por ejemplo, en el proceso de filtraje, algunos píxeles dentro de una ventana a menudo provienen de regiones con diferentes distribuciones subyacentes. Los métodos tradicionales de estimación robusta, basados en funciones de robustez simétricas, asumen que la distribución es simétrica, pero cuando la distribución de los datos es asimétrica tales métodos producen estimadores sesgados. Las funciones de influencia empírica son calculadas y, basada en esta información, esta tesis propone el Asymmetric M-estimator (AM-estimator), un M-estimador con funciones redescendentes asimétricas.
Many applications consecrate the use of asymmetric distributions to fit real data, and practical situations often require robust parametric inference. This thesis presents the derivation of M-estimators with asymmetric influence functions, motivated by the go distribution. This law, based on the multiplicative model for speckled imagery, can be highly skewed and maximum likelihood estimation can be severely hampered by small percentages of outliers. These outliers appear mainly because the hypothesis of independence and equal distribution of observations are seldom satisfied in practice; for instance, in the process of filtering, some pixels within a window frequently come from regions with different underlying distributions. Traditional robust estimation methods, based on symmetric robustifying functions, assume that the distribution is symmetric, but when the data distribution is asymmetric these methods yield biased estimators. Empirical influence functions for maximum likelihood estimators are computed, and based on this information this thesis propose the Asymmetric Mestimator (AM-estimator), an M-estimator with asymmetric redescencling functions.
Many applications consecrate the use of asymmetric distributions to fit real data, and practical situations often require robust parametric inference. This thesis presents the derivation of M-estimators with asymmetric influence functions, motivated by the go distribution. This law, based on the multiplicative model for speckled imagery, can be highly skewed and maximum likelihood estimation can be severely hampered by small percentages of outliers. These outliers appear mainly because the hypothesis of independence and equal distribution of observations are seldom satisfied in practice; for instance, in the process of filtering, some pixels within a window frequently come from regions with different underlying distributions. Traditional robust estimation methods, based on symmetric robustifying functions, assume that the distribution is symmetric, but when the data distribution is asymmetric these methods yield biased estimators. Empirical influence functions for maximum likelihood estimators are computed, and based on this information this thesis propose the Asymmetric Mestimator (AM-estimator), an M-estimator with asymmetric redescencling functions.
Description
Keywords
Procesamiento de imagen, Distribuciones asimétricas, Funciones de influencia
