Unsupervised Partitioning into Homogeneous Regions of High Resolution SAR Images using Fisher Distributions. - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2008

Unsupervised Partitioning into Homogeneous Regions of High Resolution SAR Images using Fisher Distributions.

Résumé

A new partitioning technique into homogeneous regions for High Resolution SAR images is proposed. In order to take into account the speckle noise present in such images, the intensity fluctuations are modeled with Fisher probability distribution functions (pdf) rather than more usual Gamma pdf. The partitioning is then obtained by minimizing the stochastic complexity of the image, i.e. a criterion without any tuning parameter which allows one to automatically determine the number of homogeneous regions, their position and the regularity of their contours. This technique is then illustrated on high resolution SAR images.
Fichier non déposé

Dates et versions

hal-00349869 , version 1 (05-01-2009)

Identifiants

  • HAL Id : hal-00349869 , version 1

Citer

Frédéric Galland, Jean-Marie Nicolas, Muriel Roche, Hélène Sportouche, Florence Tupin, et al.. Unsupervised Partitioning into Homogeneous Regions of High Resolution SAR Images using Fisher Distributions.. Workshop CNES/DLR, Oct 2008, Toulouse, France. ⟨hal-00349869⟩
118 Consultations
0 Téléchargements

Partager

Gmail Facebook X LinkedIn More