Exploration d'une nouvelle méthode de classification supervisée et non supervisée : application à l'analyse d'image

Abstract

The objective of this writing is to apply some classification methods to image analysis. Varied supervised and unsupervised classification technics, general or specific for image analysis, are presented to end in a new method. This one is the subject of François Roland's thesis. Its originality lies in the definition of a new dissimilarity measure. The algorithm is first tested on "simple" data sets, afterwards on an image. Finally the results are interpreted and development perspectives are proposed.

Date of Award	2006
Original language	French
Supervisor	Jean Paul Rasson (Supervisor), Marcel Remon (Jury) & Andre Hardy (Jury)