Hi Javier,
It means that the Moving and Fixed images that you are providing
as so close that the optimizer consider the current position to
be the best one and then skips some of the initialization required
for an actual optimization.
Many users tend to test this registration method by providing two
equal images with the expectation that the registration method
will return a null deformation field. They end up encountering
the problem that you are reporting.
You should simply feed this program with a realistic case of
deformation. If you need to 'fabricate' a deformaed image for
testing purposes you can use the two following programs
Insight/Examples/Registration/
BSplineWarping1.cxx
ThinPlateSplineWarp.cxx
LandmarkWarping2.cxx
You could also start by using the same images that we use for
regression testing of DeformableRegistration6 every night in
about 30 different platforms:
http://www.itk.org/Testing/Dashboard/20050402-0500-Nightly/TestDetail/__Examples_Registration_DeformableRegistration6Test.html
These testing images are:
Insight/Examples/Data/BrainProtonDensitySliceBorder20.png
Insight/Examples/Data/BrainProtonDensitySliceBSplined10.png
The second image was 'created' using the BSplineWarping1.cxx
program that we mention above.
Regards,
Luis
--------------------------
javier silva bravo wrote:
>
> C:\Documents and Settings\aldrin\Mis documentos\Imagenes\Par
> 1>DeformableRegistr
> ation6.exe Image-1.bmp Image-100.bmp ImagenSalida6.png
> DiferenciaSalida6.png Dif
> erenciaAReg.png deformacion.mha
> Starting Registration with low resolution transform
> Starting Registration with high resolution transform
> E:\InsightToolkit-2.0.0\Utilities\vxl\core\vnl/vnl_vector.h:161
> assertion failur
> e 'i<size()'
>