Skip to Main content Skip to Navigation

Contribution à la caractérisation multi-échelle de l'anisotropie des images texturées

Abstract : The thesis deals with the characterisation of textured image anisotropy. In the case of complex textures, i.e composed of both microscopic and microscopic structural elements, the author demonstrates that the measurement of either dominant direction or texture anisotropy strongly depends on the scale used for the observation.
The first chapter gives definitions for texture and for observation scale. Some popular approaches allowing texture characterisation are presented and their ability to take into account directionality at different scales is discussed.
The second chapter presents classical methods for dominant texture orientation estimation. A new local operator is proposed : the Directional Mean Vector. It is based on local characteristics of the picture and can be computed for all observation scales. The performances obtained with this operator are evaluated with synthetic and natural textures.
The third chapter introduces a new estimator, called Iso, for anisotropy. It is based on the computation of local differences of the Directional Mean Vectors obtained from a picture at a given observation scale. The performances of Iso are compared to those obtained with some classical dispersion indicators for directional data.
The last chapter is devoted to the estimation of anisotropy for complex textures versus the observation scale. A new model for complex texture is proposed and the behaviour of Iso indicator is studied for this model. Thereafter, it is shown that the evolution of Iso versus the scale of observation is obeying a law which characterises simultaneously the anisotropy of the texture and the size of the textural elements that form the picture, whether they are macroscopic or microscopic. Lastly, Iso is computed at different scales for synthetic textures, for Brodatz textures and for natural textures of composite materials obtained from electronic microscope images.
Complete list of metadata

Cited literature [87 references]  Display  Hide  Download
Contributor : Christian Germain Connect in order to contact the contributor
Submitted on : Monday, August 6, 2007 - 6:15:33 PM
Last modification on : Monday, November 26, 2018 - 1:30:05 PM
Long-term archiving on: : Friday, April 9, 2010 - 12:23:31 AM


  • HAL Id : tel-00166497, version 1


Christian Germain. Contribution à la caractérisation multi-échelle de l'anisotropie des images texturées. Traitement du signal et de l'image [eess.SP]. Université Sciences et Technologies - Bordeaux I, 1997. Français. ⟨tel-00166497⟩



Record views


Files downloads