Traitement statistique d'images hyperspectrales pour la détection d'objets diffus : application aux données astronomiques du spectro-imageur MUSE

Abstract : We study the detection and segmentation problems in extremely noised images. The main application of these works is the detection of large-scale structures in MUSE astronomical hyperspectral images, namely haloes (localized and homogenous in images) and filaments (anisotropic large-scale structures). First, we study the hypothesis-testing detection in hyperspectral images, based on spatial and spectral shape constraints as well as similarity constraints. Then, we introduce a pairwise Markov field model which allows the formulation of the detection problem as a special case of the segmentation problem, while introducing a Markovian prior on the result. Next, in order to model oriented structures in images, we propose a triplet Markov field model allowing the joint segmentation of orientations and classes in images. Finally, we study the modelling of large-scale structures in images by introducing a triplet Markov tree model handling inter-resolution dependancy jointly with homogeneity within resolutions. The two latter models were introduced in the general framework of image segmentation. Each model was validated with respect to its alternatives, then all models were compared on synthetic data in the context of detection within astronomical hyperspectral images. Finally, this document presents the analysis of the results on real MUSE images.
Liste complète des métadonnées

Cited literature [163 references]  Display  Hide  Download

https://hal.archives-ouvertes.fr/tel-01619825
Contributor : Jean-Baptiste Courbot <>
Submitted on : Tuesday, November 7, 2017 - 4:34:49 PM
Last modification on : Friday, April 5, 2019 - 8:04:59 PM
Document(s) archivé(s) le : Thursday, February 8, 2018 - 2:11:32 PM

File

Courbot_Jean-Baptiste_2017_ED2...
Files produced by the author(s)

Identifiers

  • HAL Id : tel-01619825, version 2

Citation

Jean-Baptiste Courbot. Traitement statistique d'images hyperspectrales pour la détection d'objets diffus : application aux données astronomiques du spectro-imageur MUSE. Traitement du signal et de l'image. Université de Strasbourg, 2017. Français. ⟨tel-01619825v2⟩

Share

Metrics

Record views

581

Files downloads

503