[Insight-users] setting parameters for GaussianBlurImageFunction

[Phillip Cheng]

Hi Luis,

> The way to solve the warning is not to increase
> the tolerance on the error, but to increase the
> size (in pixels) of the kernel.

Yes, but as you point out, large kernels will increase computation time
(especially since I'm working with 3D images).  I think a tolerance of 0.001
will result in kernel sizes that are too large for my current needs.

Incidentally, it's not documented what units SetExtent() is supposed to
take.  From the code it looks like it's in multiples of sigma (rather than
units of pixels or distance)?

> Just out of curiosity:
> What is your criteria for selecting the sigma at each pixel ?

The sigma I'm using depends on corresponding neighborhood information from a
segmented registered image.  This is part of some exploratory research for
partial volume correction on PET images.  The exact method of determining
sigma is part of the research question...

Phillip