Skip to Main content Skip to Navigation
Theses

Parameter Estimation and Modeling of High Resolution Synthetic Aperutre Radar Data

Abstract : The thesis is approaching this problematic by statistical modeling and Bayesian inference for complex SAR image analysis. The Tikhonov regularization method is applied for image restoration because it allows to reformulate the ill-posed image estimation problem into a well-posed problem by the selection of a convex function. It allows to use the required image and prior models and to find the Maximum A Posteriori (MAP) estimate solution, exploiting the connection to the Bayesian framework. Furthermore it allows the optimization to be performed on complex-valued data and to include the system impulse response which has to be included to correctly model the SAR image. The use of the Rate Distortion for model selection is possible because of the connection between the mutual information and the Occam factor which permits the model selection in the first level of Bayesian inference. The model selection is applied in order to optimize the parameters of the Model Based Despeckling (MBD) algorithm for image denoising and feature extraction : the optimal average analyzing window and the optimal average model order. The method is a global approach and suits in case of large data sets because of its simplicity and fastness. The Rate Distortion based model selection is appropriate for the design of image information mining systems. The Tikhonov regularization shows to be a powerfulmethod for the regularization of complex-valued images. It is recommended in applications where the phase is required, e.g. interferometry, target analysis, because it provides an estimation of the image reflectivity while preserving the phase of the signal.
Complete list of metadatas

https://pastel.archives-ouvertes.fr/pastel-00561766
Contributor : Matteo Soccorsi <>
Submitted on : Tuesday, February 1, 2011 - 5:58:21 PM
Last modification on : Friday, July 31, 2020 - 10:44:07 AM
Long-term archiving on: : Monday, May 2, 2011 - 4:24:39 AM

Identifiers

  • HAL Id : pastel-00561766, version 1

Collections

Citation

Matteo Soccorsi. Parameter Estimation and Modeling of High Resolution Synthetic Aperutre Radar Data. Signal and Image processing. Télécom ParisTech, 2010. English. ⟨pastel-00561766⟩

Share

Metrics

Record views

440

Files downloads

491