CGAL users discussion list

[Pierre Alliez <>]

dear Qianqian,

Apparently Matlab's marching cubes algorithm does not provide guarantees 
about manifold-ness of the output mesh. I do not really know what wrong 
in your mesh but it can be anything among:
- one edge sharing more than two faces,
- one vertex having more than one umbrella,
- some duplicate faces,
- etc.

although it is always possible to fix a non-manifold mesh through 
cutting and/or removing duplicated faces - it is not a well-defined 
problem - the best is to go with a method with guarantees.

it is possible that CGAL will provide in the next release a 3D mesh 
generator which allows you to bypass the MC+decimation stage and 
generates as output a mesh with guaranteed properties.

pierre


Qianqian Fang a écrit :
> thank you Pierre for your prompt reply and help.
>
> I tried your new mesh with the cgal surface mesh simplification code. But
> strangely, it exited with the error message "CGAL error: precondition
> violation!".
>
> Indeed, I am not really worry about this particular mesh. My real goals are
> to understand the deficiency for the mesh produced by matlab's isosurface,
> and being able to repair it automatically (for any 3D volume).
>
> To explain it more, I am writing a meshing tool to create 3D tetrahedral
> mesh from any binary volumetric image, for example, a thresholded
> image from a MRI scan of a human head. So far, I have been coded
> a matlab script to do the followings:
> 1. extract the surfaces from the binary volume image using matlab's 
> isosurface
> 2. resample the surface calling CGAL's surface mesh simplification code
> 3. dump the simplified surfaces (must be closed surface, adding the 
> bounding box
> cut-out to make it close if necessary) into piecewise linear complexies 
> (see [1])
> 4. call tetgen [2] externally to produce tetrahedral mesh from the 
> boundaries
>
> tetgen seems to be working fairly robustly. The mesh simplification was 
> mostly causing
> problems. In this case, it was matlab's isosurface produced problematic 
> mesh
> base on your comments. I am seeking either alternative path to do the above
> work-flow, such as completely using CGAL as you suggested, or trying to 
> figure
> out what's wrong with matlab's isosurface (this is indeed preferred, as 
> it is already
> working for most of my cases, except for one or two as I reported earlier).
>
> Comparing your modified mesh, I saw you added some duplicated nodes,
> can you share me with more details on why the original matlab's mesh 
> failed?
> and is there a way to fix it automatically?
>
> If using CGAL exclusively for my above workflow, are there examples to 
> produce
> surface mesh from a binary volumetric image? If it is not difficult to
> deal with this in CGAL, I might consider the alternative path instead.
>
> thank you
>
> Qianqian
>
> [1] http://tetgen.berlios.de/plc.html
> [2] http://tetgen.berlios.de/
>
> Pierre Alliez wrote:
> > apparently the matlab marching cubes is just bad - there exists some 
> > better marching cubes which guarantee 2-manifold - the attached mesh 
> > is better I think - but my fix was manual.
> >
> >
> > perhaps you could start considering CGAL surface mesh generator?
> >
> >
> >
> > Qianqian Fang a écrit :
> > > the modified mesh works ok with the surface mesh simplification 
> > > program. However, the results after the operation present edges. We 
> > > are expecting a closed surface for this mesh before and after the 
> > > simplification.
> > >
> > > I apologize that I don't have much experience with CGAL and can not 
> > > put up with a "incremental builder" quickly to find out the error 
> > > message. I am really curious to know what are the problems with the 
> > > original mesh (it was produced from isosurface in matlab). Any tips 
> > > for fixing this type of problems is also appreciated.
> > >
> > > Qianqian