DELEDALLE, Charles-Alban; DENIS, L.; TUPIN, Florence

Métadonnées

Afficher la notice complète

Licence d’utilisation du document

DELEDALLE, Charles-Alban
Institut de Mathématiques de Bordeaux [IMB]

DENIS, L.
Laboratoire Hubert Curien [LabHC]

TUPIN, Florence
Image, Modélisation, Analyse, GEométrie, Synthèse [IMAGES]
Département Images, Données, Signal [IDS]

Langue

Communication dans un congrès

Ce document a été publié dans

IGARSS, IGARSS, IGARSS, 2018-07, Valencia. 2018-07

Résumé en anglais

Speckle reduction is a long-standing topic in SAR data pro-cessing. Continuous progress made in the field of imagedenoising fuels the development of methods dedicated tospeckle in SAR images. Adaptation of a denoising techniqueto the specific statistical nature of speckle presents variablelevels of difficulty. It is well known that the logarithm trans-form maps the intrinsically multiplicative speckle into anadditive and stationary component, thereby paving the wayto the application of general-purpose image denoising meth-ods to SAR intensity images. Multi-channel SAR imagessuch as obtained in interferometric (InSAR) or polarimetric(PolSAR) configurations are much more challenging. Thispaper describes MuLoG, a generic approach for mappinga multi-channel SAR image into real-valued images withan additive speckle component that has a variance approxi-mately constant. With this approach, general-purpose imagedenoising algorithms can be readily applied to restore InSARor PolSAR data. In particular, we show how recent denois-ing methods based on deep convolutional neural networkslead to state-of-the art results when embedded with MuLoGframework.< Réduire

Mots clés en anglais

SAR

speckle

interferometry

polarimetry

multiplicative noise

URI

https://oskar-bordeaux.fr/handle/20.500.12278/193235

Project ANR

Analyse de surfaces urbaines par tomographie SAR - ANR-15-ASTR-0002

Origine

Importé de hal

Unités de recherche

Institut de Mathématiques de Bordeaux (IMB) - UMR 5251