[Insight-users] itkLinearInterpolateImageFunction

[Luis Ibanez]

Hi Anja,


Your concern actually applies equally to linear and non-linear registration.

It is commonly the case that when you map the grid point of the
Fixed image into the coordinate system of the Moving image, the
points will land in non-grid positions of the Moving image and
therefore their values will have to be interpolated.

The type of Transform that maps the point coordinates from the
Fixed image to the Moving image is irrelevant in this case. A
simple rotation, or any subpixel translation will already produce
this effect.


The role of the interpolator is to simply restore continuity in
a functions (Moving image) that has been previously discretized.

You could use any Interpolator in combination with any Transform.

The choice of an interpolator is rather driven by a compromise
between the quality that you want for the interpolated values
and the amount of computation that you are willing to invest in
order to get those values.


     Regards,


         Luis


-----------------
Anja Ende wrote:
> Hello all,
> 
> I was looking at the implementation of the linear interpolation filter
> in itk and found it in itkLinearInterpolateImageFunction.txx.
> 
> This class assumes that the image sampled points lie on a regular
> grid, right? What about interpolation when doing non-rigid
> registration? The index points generated from the registration process
> would lie on an irregular grid, right? What interpolation classes
> could we use then?