[vtkusers] Moving Least Squares Surface Reconstruction

[David Doria]

Is anyone familiar with MLS surface reconstruction?

Section 2.3.3 of Point Primitives for Interactive Modeling and
Processing of 3D Geometry (ftp://ftp.inf.ethz.ch/doc/diss/th15134.pdf)
says (obviously) that the normals must be consistently oriented to
extract the MLS surface as the zero level set. The only way I could
think to do this is:

1) compute all of the projected points you will need for the signed
distance field (project all of the points in the uniform 3D grid)
2) compute the normals at the projected points
3) orient the normals on the projected points (using the EMST method)
4) then construct the 3D uniform signed distance field

Does that sound like a reasonable procedure?

(Is there a better place to ask a question like this than the VTK
mailing list? It will soon be VTK relevant because I'll write it as a
VTK filter :) ).

Thanks,

David