[vtkusers] Scalar Value in MC

[Erik Sohns]

Hi all,

I'm trying to contour segmented volumes using marching cubes. Initially,
I'd taken the a straightfoward approach, 
setting the scalar value to the actual class value (i.e. 1024) I'd
wanted to visualize, having masked the unwanted classes aforehand. This
seems to work, although I get a rather stepped surface. The steps occur
between voxel boundaries.

Now, the book gives an example using a slightly different technique,
where half the contour value (i.e. 511.5) is used. This 'shifts' the
generated surface by half a voxelsize, producing slightly more
primitives in my case. In this case, the steppings occur at mid-voxel
height (as expected). As these generated models are not equivalent, I
was wondering which is the better approach respective model accuracy.

Any ideas ?

Greetings,
Erik