CGAL users discussion list

["Amir Vaxman" <>]

Hello,

 

My input is a segmented, 2-manifold, surface mesh, and I would like to create a tet-mesh of the inscribed volume, which will subscribe to that segmentation in the sense that every inner tet is labeld as the surface triangle that is closest to it (say, closest to its four points). This should result in a tet mesh of the different segmentation labels, in which the boundaries between the different labels is an approximate subset of the medial axis of the surface (and, of course, the surface itself).

 

I tried two approaches:

 

1. Sampling a grid of points for closest triangle to each point, in order to label them (and in\out of the mesh), and then using a multi-label grid mesh generation scheme. The problem is that this sampling is very costly and that complicated surfaces need a very high resolution in order to be reproduced reliably.

 

2. Generating a segmentation-unaware mesh (with standard surface-to-mesh tools), and then labeling the tets according to distance. The major problem here is that because the meshing is segmentation-blind, then the inner tets do not produce a fine separation surface between the labels inside the mesh.

 

Does anyone have a better solution for that problem (or an algorithm\code) that does the segmented-surface to segmented-volume directly?

 

Thanks,

Amir.