[vtkusers] vtkThinPlateSplineTransform limits

Tim Hutton T.Hutton at eastman.ucl.ac.uk
Fri Jun 21 06:46:20 EDT 2002

Good question. In terms of computational cost, the expensive bit is 
decomposing an N+D+1 x N+D+1 matrix, where N is the number of points and D 
is the number of dimensions (2 or 3) and decomposition is an O(N^3) 
operation. If you're using more than, say, 50 or 100 points then it can 
start to slow down. While I haven't pushed vtkTPS to see when it will 
break, I think there is no intrinsic limit other than the amount of memory 
available and your patience.

One way to get around having to do one big TPS warp would be to divide your 
object into different parts if this is possible, and do a smaller TPS on each.

On a related note, vtkMath uses the Jacobi method for eigen-decomposition - 
there are faster ways that you could use if this would help (and contribute 
to VTK if they do).


At 11:51 21/06/2002 +0200, Steve Boyd wrote:
>Does somebody know the practical limitations of the
>vtkThinPlateSplineTransform in terms of the number of target and source
>points it can typically handle?  Are there ways to get around
>limitations related to the number of points used?
>Steven Boyd, PhD
>Institute for Biomedical Engineering
>ETH and University Zuerich
>Moussonstrasse 18
>8044 Zuerich, Switzerland
>tel. +41.1.632.4591  fax. +41.1.632.1214  boyd at biomed.ee.ethz.ch
>This is the private VTK discussion list.
>Please keep messages on-topic. Check the FAQ at: 
>Follow this link to subscribe/unsubscribe:

More information about the vtkusers mailing list