offres d'emploi formations actualités contact accès annuaire intranet
Séminaires >

A proposed normalized B-spline density estimator and it application in unsupervised statistical image segmentation.

Atizez Hadrich, Faculté des Sciences de Sfax, Laboratoire de probabilités et statistique

jeudi 24 mai 2012 à 13h30

salle B014


This talk describes a new density estimation method of distribution mixture based on B-spline density estimator with application to unsupervised statistical image segmentation. The proposed normalized B-spline density estimator overcomes the situation where the orthogonal series density estimator is not a probability density function (pdf). This estimator is competitive and bears a striking resemblance to the orthogonal series density estimator. We introduce the proposed estimator for estimating the mixture density. The application of suggested approach in unsupervised statistical image segmentation does not make heavy assumptions on the shape of the gray level image pixels distribution.