LIP6 2000/021:
THÈSE de DOCTORAT de l'UNIVERSITÉ PARIS 6 LIP6 /
LIP6
research reports
150 pages - Décembre/December 1999 -
French document.
Get it : 2697 Ko /Kb
Contact : par mail / e-mail
Thème/Team: Apprentissage et Acquisition de Connaissances
Titre français : Représentation de la variabilité dans le traitement d'images flou
Titre anglais : Representation of variability in fuzzy image processing
Abstract : One of the most important factors, which limit the performance of classifiers in many image processing tasks, is the variability. In this thesis, we will propose a new method to reduce the effect of variability in a fuzzy classification system for image processing by using the context in the image.
The method consists of the following steps:
- Parameterization of the histograms of all the attributes that are calculated on the images for the classification task which gives a compact description with a small number of parameters.
- Conctruction of a prototype which includes the variability and the interactions between the parameters using the learning database.
- Adaptation of the model to the current case by adapting the parameters of the prototype.
The proposed method is applied to a synthetic data base and to a database of mammography images for the detection of dense lesions.
Key-words : Fuzzy Classification, Inductive Learning, Variability, Mixture Models, EM Algorithm, Regression, Prototypes, Markov Models, Mammography, Computer Aided Detection
Publications internes LIP6 2000 / LIP6 research reports 2000