Skip to Main content Skip to Navigation

Détection des galaxies à faible brillance de surface et segmentation hyperspectrale dans le cadre de l'observatoire virtuel

Abstract : Technological progress in astronomical instrumentation raise various
issues. Thanks to the growing sensor sensitivity,
monospectral imagery make it possible to discover objects which were
previously impossible to detect. The development of multispectral sensors
yields extremely valuable data. Nevertheless interpretation and
processing of such images remain tricky for the astronomical
community. Within the framework of this thesis we propose a set of
methods that make the interpretation process easier for the
astronomer. We introduce a new fuzzy segmentation method based on Markov
fields allowing to take into account the specificities of astronomical
objects : fuzzy boundaries and diffuse objects. A
fuzzy pixel of the segmentation map thus belongs to one or two hard
classes depending on a certain membership level. We also
propose a new method for the detection of Low Surface Brightness (LSB)
galaxy based on a quadtree Markovian segmentation.
This segmentation allows to highlight the LSB within the observation
background through an accurate estimation of the noise
statistics contained in the acquisition. A set of selection steps is
then carried out to determine if the detected object is a LSB.
We then introduce two multispectral images visualization methods
allowing to synthetize the information contained by all the
observation bands in a colored composition in the HSV color
space (Hue Saturation Value). Finally we propose a new
segmentation method of hyperspectral data cubes based on a spectral
discrimination and on a spatial regularization of the
segmentation map obtained thanks to a quadtree segmentation. These
methods are validated on astronomical images and led to a fruitful cooperation
between computer vision community and astronomical community. Furthermore
two methods have been validated on remote sensing
observation for which some specific issues remain common.
Complete list of metadata

Cited literature [81 references]  Display  Hide  Download
Contributor : Matthieu Petremand <>
Submitted on : Wednesday, June 6, 2007 - 9:30:29 AM
Last modification on : Thursday, April 23, 2020 - 2:26:30 PM
Long-term archiving on: : Friday, September 21, 2012 - 4:15:16 PM


  • HAL Id : tel-00152065, version 1



Matthieu Petremand. Détection des galaxies à faible brillance de surface et segmentation hyperspectrale dans le cadre de l'observatoire virtuel. Traitement du signal et de l'image [eess.SP]. Université Louis Pasteur - Strasbourg I, 2006. Français. ⟨tel-00152065⟩



Record views


Files downloads