[Insight-users] RecursiveGaussianImageFilter

[Gunnar Farneback]

In February Luis wrote:
> The good news:
> 
>      We were able to reproduce the effect that you
>      reported. That is, the values of the derivatives
>      change when you add a constant to the input image.
> 
> 
> The bad news:
> 
>       We still have to figure out why ...
>       and...  how to fix it.

I have figured out both why and how. A revised implementation and some
further comments are available at
http://lmi.bwh.harvard.edu/~gunnar/itk_recursive_gaussian.html
Please test it.

> My guess is that this effect is related to the
> fact that this filter is not just doing derivatives
> but smoothing followed by derivatives.

No, the main problem was that the filters were simply not designed to
have zero response to constant signals. The new implementation
includes a revised filter design.

/Gunnar