Hierarchical 3D Segmentation Using Connected Face Structure

Sébastien Delest (1) , Romuald Boné (1) , Hubert Cardot (1)

1 LIFAT - Laboratoire d'Informatique Fondamentale et Appliquée de Tours

This paper describes a new approach for watershed segmentation on triangular mesh. Common watershed approaches use descending or flooding processes on a connected vertex structure. Watershed transformation is not related to a structure in particular: it only requires connected elements and a height function. This method is widely used on 2D, 3D images as well as on 3D meshes. Here, a connected face structure is implemented and adapted to the segmentation process. A connected face structure offers a different kind of curvature information and neighborhood. The waterfall algorithm based on the minimum spanning tree is used to compute the merging. Several segmentation schemes are built from the waterfall and they can easily be browsed by the user.

Domaines: Traitement des images [eess.IV]; Traitement du texte et du document; Vision par ordinateur et reconnaissance de formes [cs.CV]

Sébastien Delest, Romuald Boné, Hubert Cardot. Hierarchical 3D Segmentation Using Connected Face Structure. International Journal for Computational Vision and Biomechanics, 2008, 1 (2), pp.1-13. ⟨hal-01022567⟩