[vtkusers] Holes and inverted normal after vtkQuadricClustering

[Thiago Franco Moraes]

Hi all,

I've been using vtkMarchingCubes to generate a vtkPolydata, but
instead of generating polydata from the whole imagedata at once I've
been splitting the imagedata and generating polydata from each
imagedata part using the vtkMarchingCubes and in the final I join all
the generated polydatas in one polydata using vtkAppendPolydata. To
each polydata part I've been using vtkQuadricClustering to decimate
them before join them. After the joining all polydata a
vtkCleanPolydata and vtkPolydataNormal is used. So my pipeline is like
this:

vtkImagedata -> vtkMarchingCubes -> vtkQuadricClustering
vtkImagedata -> vtkMarchingCubes -> vtkQuadricClustering \
....
                 -> vtkAppendPolydata -> vtkCleanPolydata ->
vtkPolydataNormals
vtkImagedata -> vtkMarchingCubes -> vtkQuadricClustering/


But in final the vtkPolydata has a lots of little holes ( a example in
[1]), if I export to STL and open in Meshlab, for example, the
polydata has a some parts with inverted normals [2]. If I don't use
vtkQuadricClustering those problem doesn't occurs. Apparently
vtkQuadricClustering is what generate those problems. In Marching I'm
setting it to compute the normals, gradients and scalars. In
vtkQuadricClustering I'm setting this way:

       decimation.SetDivisionOrigin(0, 0, 0)
       decimation.SetDivisionSpacing(imagedata.GetSpacing())
       decimation.SetFeaturePointsAngle(120) # I teste with others
angles too (80, 30, 45 ...)
       decimation.UseFeaturePointsOn()
       decimation.UseFeatureEdgesOn()
       decimation.UseInputPointsOn()
       decimation.UseInternalTrianglesOn()
       decimation.CopyCellDataOn()

Why vtkQuadricClustering is generating those holes and inverted the
normals? Is there something I can do?

Thanks!

[1] - http://i.imgur.com/kCvxX.jpg
[2] - http://i.imgur.com/YF1es.jpg