[vtkusers] optimized thin-plate-spline in VTK ?

[Liao Qinghu]

hi, 

you mean the Fast Multipole Method (FMM). there is a trial version of
the fast radial method. FastRBF

http://www.farfieldtechnology.com/products/toolbox/


On Thu, 2003-12-18 at 09:22, Khaldoun Zine El Abidine wrote:
>  
> 
> Hello everybody,
> 
>  
> 
>  
> 
>  
> 
> I am interested in an “optimized” Thin Plate Spline (TPS)
> interpolation/smoothing algorithm (in 2D). It seems to me that VTK’s
> implementation uses the direct method (i.e. one that inverts a N*N
> matrix). If that is the case, VTK’s implementation has a O (N^2)
> memory usage and a O (N^3) computational complexity, which is
> impractical for N > 5000 when execution time requirements are strict
> (no more than a few seconds on a regular PC). Algorithms in O (N log
> N) complexity exist in the literature (see for example “Fast Solution
> Of The Radial Basis Function Interpolation Equations: Domain
> Decomposition Methods, R. K. BEATSON, W. A. LIGHT, AND S. BILLINGS, in
> Society for Industrial and Applied Mathematics, Vol. 22, No. 5, pp.
> 1717–1740, 2000”, among others). Is anyone aware of a working
> implementation of this algorithm? Is it, or will it be, available in
> present/future versions of VTK ?
> 
>  
> 
>  
> 
> Thank you in advance for you help.
> 
>  
> 
>  
> 
>  
> 
> Khaldoun.
> 
>  
-- 
Qinghu Liao
Department of Computer Engineering,
Polytechnique, University of Montreal
CANADA

qing-hu.liao at polymtl.ca
qhliaok at hotmail.com