Hierarchical watersheds within the Combinatorial Pyramid framework - Archive ouverte HAL Access content directly
Conference Papers Year : 2005

Hierarchical watersheds within the Combinatorial Pyramid framework

Luc Brun
  • Function : Author
  • PersonId : 936962
Myriam Mokhtari
  • Function : Author
  • PersonId : 837761
Fernand Meyer

Abstract

Watershed is the latest tool used in mathematical morphology. The algorithms which implement the watershed transform generally produce an over segmentation which includes the right image's boundaries. Based on this last assumption, the segmentation problem turns out to be equivalent to a proper valuation of the saliency of each contour. Using such a measure, hierarchical watershed algorithms use the edge's saliency conjointly with statistical tests to Decemberimate the initial partition. On the other hand, Irregular Pyramids encode a stack of successively reduced partitions. Combinatorial Pyramids consitute the latest model of this family. Within this framework, each partition is encoded by a combinatorial map which encodes all topological relationships between regions such as multInformation Processing Letterse boundaries and inclusion relationships. Moreover, the combinatorial pyramid framework provides a direct access to the embedding of the image's boundaries. We present in this paper a hierarchical watershed algorithm based on combinatorial pyramids. Our method overcomes the problems connected to the presence of noise both within the basins and along the watershed contours.
Fichier principal
Vignette du fichier
article_luc_Brun.pdf (154.61 Ko) Télécharger le fichier
Origin : Files produced by the author(s)
Loading...

Dates and versions

hal-00126316 , version 1 (24-01-2007)

Identifiers

  • HAL Id : hal-00126316 , version 1

Cite

Luc Brun, Myriam Mokhtari, Fernand Meyer. Hierarchical watersheds within the Combinatorial Pyramid framework. Discrete Geometry for Computer Imagery, Apr 2005, Poitiers, France. pp.34-44. ⟨hal-00126316⟩
4717 View
175 Download

Share

Gmail Facebook X LinkedIn More