Skip to Main content Skip to Navigation
Theses

Segmentation d'images et morphologie mathématique

Abstract : Image segmentation by mathematical morphology is a mothodology based on the notions of watershed and homotopy modification. These tools are built starting from elementary morphological transformations which are presented in the first part of this thesis. These basic transformations are the morphological operations applied to grey-tone images and, in particular, the thinning and thickening operators together with the geodesic transformations. These tools lead to the design of more sophisticated transforms. Among them, the morphological gradient and its regularization,and the watershed transform. The latter transformation is introduced and itsrelationship with the geodesic operators and the homotopic thickenings is emphasized. Then various watershed algorithms are presented using the skeleton of a function and the representation of grey-tone images as a graph of arrows.
The second part is devoted to the use of these tools. A fair segmentation can be obtained when we use markers of the regions to be extracted to change the homotopy. These tools are also used for more complex segmentation. An image hierarchy is defined through the watershed transform.This hierarchy allows the segmentation of images where region marking is more difficult. Another example is given showing the difficulties as well as the advantages of this methodology.
Document type :
Theses
Complete list of metadatas

Cited literature [13 references]  Display  Hide  Download

https://pastel.archives-ouvertes.fr/tel-00108290
Contributor : Serge Beucher <>
Submitted on : Friday, October 20, 2006 - 12:10:55 PM
Last modification on : Thursday, September 24, 2020 - 4:38:02 PM
Long-term archiving on: : Tuesday, April 6, 2010 - 8:16:22 PM

Identifiers

  • HAL Id : tel-00108290, version 1

Citation

Serge Beucher. Segmentation d'images et morphologie mathématique. Mathématiques [math]. École Nationale Supérieure des Mines de Paris, 1990. Français. ⟨tel-00108290⟩

Share

Metrics

Record views

1799

Files downloads

14926