[Insight-developers] Parameterization of surfaces meshes

[Luca Antiga]

Hi Alex,
  very interesting. I'll be happy to take a look at the submission  
when it comes out.

By the way, I'm using Laspack (http://www.mgnet.org/mgnet/Codes/ 
laspack/)
as a sparse linear system solver library (it provides all major  
preconditioned iterative
Krylov solvers).
It's a bit old (ten years) but it works, it's cross-platform and self- 
contained. The only
issue is the current licence, but it's maybe worth to ask the author.
I have CMake-d version, I can send it to you if you want to try it out.

Regards

Luca



On Oct 18, 2007, at 11:24 PM, Alexandre GOUAILLARD wrote:

> Hi guys,
>
> Good news, our code for parameterization of surface meshes is now  
> ready to
> transfert to ITK.
>
> Spherical (possibly conformal) parameterization has been used  
> intensively
> for brain mapping, but is unfortunatly limited to zenus zero  
> surfaces. A
> conformal flatennign filter is actually in /Review, but it uses a  
> spherical
> parameterization to begin with, so it suffers from the same  
> limitations.
>
> We have implemented most of the (border-fixed) parameterization you  
> can find
> out there in a framework. We provide a simple algorithm that can  
> extract a
> "cut graph" from surfaces meshes of ANY genus. Cutting the surface  
> open alng
> this cut graph allows you to parameterize it. You can choose what  
> kind of
> border transform you want both in term of shape (usually disk or
> square-shaped domain, but any convex domain will do) and in term of  
> energy
> minimization. Same goes for the parametrization itself, and finally  
> the code
> is templated over the matrix solver.
>
> This last point is quite important, as different solvers will give  
> you very
> different speed depending on your specific case. It is also  
> important, as
> only vnl, which is quite a slow solver, is usable natively from  
> ITK, so you
> are more likely to use anything else. Templating over the solver  
> let you do
> so without modifying the algorithm itself (a-la C-GAL).
>
> The firsts results illustrate our point:
> Parameterization of one surface mesh with vnl:   6.1 s
> Parameterization of the same mesh with  TAUCS:   0.27s
>
> Expect a submission of a roughly 4 pages paper around end of  
> november 2007.
>
> We are also working out the best way (integration vs. compilation  
> flag) of
> using TAUCS with ITK. TAUCS copyright owner agreed on a BSD  
> licence, but we
> still need to check if the librairies TAUCS is build onto have  
> suitable
> licence, and if it would make sens for ITK to intergrate it.
>
> If you have any comment, any idea, or if you wanna give a look at  
> the code
> beforehands, don't hesitate to send me an e-mail. We are sending  
> this e-mail
> to gather attention and thus have people reviewing the paper quickly,
> hopefully reducing the amount of time needed to transfer the code into
> /Review.
>
> The work is based on previous publications (see below). This  
> version is
> written on top of itkQuadEdgeMesh, a new datasrtucture dedicated to  
> surface
> meshes. It has been enterely re-written by Arnaud Gelas (BioImaging  
> Lab,
> Singapore).
>
> Alex.
>
> A. GOUAILLARD, A. GELAS,  S. VALETTE,  E. BOIX and R. PROST,
> “Curvature-based Adaptive Remeshing for Wavelet-Based Multires. 3D  
> Meshes”.
> In Proc. of International Conference on Image Processing ICIP’05,   
> Genova,
> September 11th-14th 2005. Vol. 1,  pp 1033-1036.
>
> A. GOUAILLARD, A. GELAS, T. KANAI, R. PROST,
> « Image Processing and Computer Graphics: Illustrated  
> Complementaries ».
> JSF 2004, 4-5 Nov., Yoyogi, Tokyo.
>
> A. GOUAILLARD, A. GELAS, S. VALETTE, E. BOIX, R. PROST,
> “Remeshing Algorithm for Multiresolution Prior Model in Segmentation."
> In Proc. of International Conference on Image Processing, ICIP‘04,
> Singapore, October 24~27 October 2003, pp. 2753-2756.
>
>
>
> _______________________________________________
> Insight-developers mailing list
> Insight-developers at itk.org
> http://www.itk.org/mailman/listinfo/insight-developers