[Insight-users] Interpoloation in MattesMutualInformation

[Luis Ibanez]

Hi Eve,


1) The Mattes' MI implementation in ITK will
    use the interpolator that you provide. It
    is not limited to be BSplines, as the choice
    in the original paper.


2) If you decide to use a BSpline interpolator
    then the Mattes' MI implementation will take
    advantage of the properties of this BSpline.

    See lines 1232-1241 of the file:

    Insight/Code/Algorithms/
      itkMattesMutualInformationImageToImageMetric.txx


3) Yes the use of the interpolator is independent
    of the use of BSpline for performing Parzen windowing
    on the estimation of the joint histogram.

    The BSplines used for Parzen windowing are created
    in lines 309-310. Regardless of what type of interpolator
    you use.


4) Note also that this metric has a special case when the
    Transform itself is a BSpline deformable transform.



   Regards,


     Luis


============
Eve wrote:
> Hi ITK-Users,
> 
> I am confused as to the uses of B-splines in Mattes' implementation of MI
> interpolation. In Mattes' paper, it seems that B-splines are also used to
> resample the moving image (section II.A in his paper); in ITK, it seems that
> linear-interpolation (or any set through registration->SetInterpolation( ..
> )) is used by default, true? This use of B-Splines is totally different from
> the B-Splines in the Parzen window estimate, correct?
> 
> Thanks,
> Eve