Apport d'une nouvelle mesure de dissimilarité : un algorithme de segmentation basé sur les graphes

Abstract

This work addresses the problem of segmenting an image into regions. We first define a predicate for measuring the evidence for a boundary between two regions using a graph-based representation of the image. We then develop an efficient segmentation algorithm based on this predicate. \\ We apply the algorithm to image segmentation using two different dissimilarity between pixels : the Euclidean distance and a new empirical distribution-based dissimilarity. We illustrate the results so that we can underline the contribution of this new dissimilarity to image segmentation.

Date of Award	2007
Original language	French
Supervisor	Jean Paul Rasson (Supervisor), Andre Hardy (Jury), Marcel Remon (Jury) & François Roland (Jury)