[vtkusers] Displaying Every Voxel in a 3D Image

[Valerie Coffman]

Dear VTK Users and Developers,

I am writing about a problem with displaying *all* of the voxels in a 3D 
image.  A related problem was discussed in this thread:
http://public.kitware.com/pipermail/vtkusers/2002-July/061952.html

I am using a vtkImageData object with the 
vtkFixedPointVolumeRayCastMapper.  The problem is that an X by Y by Z 
image is displayed as an X-1 by Y-1 by Z-1 image.  Clearly, this is not 
the behavior a user would expect.  (I'm aware that with other mappers, 
instead of placing the sample point at the lower left corner , the 
voxels are centered around the sample points -- all voxels are 
displayed, but the voxels on the boundaries are cut.)  We and our users 
expect to load a 3D image and display every voxel as an equally sized 
box.  In order to achieve this with the current VTK code, we have to 
either pad our images on three sides or copy the 
vtkFixedPointVolumeRayCastMapper code and modify a few of the 
functions.  Both of these options are messy hacks.

I understand the argument, made by Ken Martin in the thread I linked to 
above, that images are defined between their sample points.  However I 
feel strongly that the user should be able to display every voxel in its 
entirety without resorting to messy hacks.  Does anyone have a 
suggestion for a cleaner work around?  One possible solution is to add 
an option to the vtkFixedPointVolumeRayCastMapper that allows the users 
to choose the bounds of the vtkVolume independently of the bounds of the 
vtkImageData such that the bounds contain all of the voxels.  Would the 
vtk developers consider adding this functionality?  I am willing to 
submit the code.

Thank you,
Valerie