[vtkusers] How to average different meshes (from surface points)?

[Arnaud GELAS]

Hi Lynx,

Most of surface reconstruction methods extrapolate wherever samples are 
missing. I assume the "surface flaps" are located in a region where data 
are missing?
In this case vtkSurfaceReconstructionFilter provides "correct" results 
in regions where you have enough samples...

Now the problem is to know whether you want to extrapolate in region 
where samples are missing or not. In some case, due to the acquisition 
process some data are missing (holes), and you know that in reality 
these holes do not exist. Is that your case? Or is this the end of the 
surface patch that you want to reconstruct?

In this last case, you need to change the bounds of the meshing part 
(marching cubes) to some bounds that naturally fit the region of interest.

I am not sure if I fully understand your problem, but hope it helps...

Arnaud


On 10/27/2009 04:04 PM, lynx.abraxas at freenet.de wrote:
> On 26/10/09 17:20:23, Arnaud GELAS wrote:
>    
>> Can you send one image of the results please?
>>      
> I've  attached  a  montage  of  SetNeighborhoodSize variing from 20 to 37. The
> colour variation has no meaning,  (I  think).  Some  incorrect  surface  flaps
> appear  more  often  than  others  but only the fitting surface (more detailed
> part) seems to be stationary (at least to some useful approximation).
> Would some kind of averaging do magick here?
> Which would be better: adding up  vtkSurfaceReconstructionFilter  output  (for
> different  SetNeighborhoodSize  values)  as  is  or  using  contour points for
> vtkGaussianSplatter average these and countur again?
>
> Many thanks.
> Lynx
>