Re: [Algorithms] Edge collapse based simplification problem
Brought to you by:
vexxed72
From: Tom F. <tom...@bl...> - 2003-02-08 13:41:45
|
The nice thing about the test for normal-flipping is that it also gives you a quality improvement. These are cases where the topology is still perfectly valid after a collapse, but that collapse was visually not a good one to make because it resulted in flipped-normal triangles. By "flipped" I mean that the normal of any of the new triangles is more than 90 degrees away from the normal of any of the old triangles. That usually removes most evils. Fortunately these cases are rare (but even one can be very visible), so it's usually enough to do the collapse, _then_ do the test, and find it's a flipper, then undo it and mark that collapse as bad. Not efficient, but easy to code, and as I say, they're rare. TomF. ----- Original Message ----- From: Joe Ante <jo...@li...> To: <gda...@li...> Sent: Friday, February 07, 2003 11:53 PM Subject: Re: [Algorithms] Edge collapse based simplification problem > Hi Peter, > > The standard way people solve this (as far as I know) is to just check for > > faces flipping after the collapse (and not allowing it if this happens.) > How exactly do I test for flipping after the collapse? > > > (I think the topological test might just be if the intersection of the 1 rings > > of two vertices on the edge that is being collapsed contains vertices that > > aren't on the face(s) that are going to be removed after the collapse the > > resulting mesh would be non-manifold. I believe hugues discusses this in his > > thesis and Edelsbruner (sp?) has papers that discuss this as well... > Ive not been able to find these papers, hoppes thesis seems to be availible > only as ppt presentation. > > > So > > basically the intersection of the 1 rings should have 4 vertices (3 if it's a > > boundary edge.) > Ah that sounds like a much simpler test, thanks. > > Are you sure that the intersection should be 4 vertices for non boundary > edges? > When I draw it on paper the 1ring intersection seems to be 3 for > non-boundary and 2 for boundary edges when the edge will become > non-manifold. > > Joe ANte > > > > ------------------------------------------------------- > This SF.NET email is sponsored by: > SourceForge Enterprise Edition + IBM + LinuxWorld = Something 2 See! > http://www.vasoftware.com > _______________________________________________ > GDAlgorithms-list mailing list > GDA...@li... > https://lists.sourceforge.net/lists/listinfo/gdalgorithms-list > Archives: > http://sourceforge.net/mailarchive/forum.php?forum_id=6188 |