[Insight-developers] Derivatives of Recursive Gaussian dx scaling?

[Ivan Macia]

Hello,

What you say makes sense. If the RecursiveGaussianImageFilter knows about
derivative order and spacing, it could perform the spatial normalization
too.

Regarding the GaussianDerivativeOperator, considering spacing or not inside
is really a design decision. In theory, the operator should not need to know
about image spacing, since this comes with the image function. However, in
practice, most of the related image functions use several operators and
setting coefficients for all of them in each image function could be
redundant and error prone. If you ask me, I would let the operator perform
the spatial normalization, unless I am missing some other implication.

It would be nice of the GaussianDerivativeOperator, and possibly related
image functions, could be part of the core of the toolkit :) (this
encourages me to submit some pending contributions).

Another side note is that kernel based computations could be greatly speed
up in the future by implementing them via GPU. I see that something similar
is done in CUDA ITK. I did not make any performance tests on the operator
and related and image functions, maybe they could be improved in that sense,
especially the image functions.

Best regards

Iván


2010/12/28 Hans Johnson <hans-johnson at uiowa.edu>

>  Brad,
>
> I would think that now is the time to do the consolidation.  Your argument
> seems correct to me.
>
> Hans
>
>
>
>
> On 12/27/10 9:43 AM, "Bradley Lowekamp" <blowekamp at mail.nih.gov> wrote:
>
> Hello,
>
> So I have been taking a very close look into the Derivative calculations of
> RecursiveGaussianImageFilter ( and related Laplacian, gradient, and Hessain
> ) along with GaussianDerivativeOperator (also related functions ). Why
> doesn't RecursiveGaussianImageFilter divide by spacing^order internally.
> Currently, the recursive Laplacian, Gradient and Hessian filters divide by
> it internally. The reason why I am asking is because currently the
> GaussianDerivativeOperator and related functions are missing this scaling
> factor. I am not sure if it should go into the Operator on in the functions,
> so I wanted to ask why this similar filter was implemented the way it was to
> assist with this design decision.
>
>
> Also what is the future of the GaussianDerivativeOperator? Normally is
> would just be a matter of moving from review into Common/BasicFitlers,
> However with the ITKv4 it may make since to do a little consolidation as
> well. The GaussianDerivativeOperator and the GaussianOperator contain a good
> bit of redundant code, so the GaussianDerivativeOperator could replace the
> later with a different set of defaults.
>
> Thanks,
> Brad
>
> * *
>
>
>
> ========================================================
>
> Bradley Lowekamp
>
> Lockheed Martin Contractor for
>
> Office of High Performance Computing and Communications
>
> National Library of Medicine
>
> blowekamp at mail.nih.gov
>
>
>
> ------------------------------
> _______________________________________________
> Powered by www.kitware.com
>
> Visit other Kitware open-source projects at
> http://www.kitware.com/opensource/opensource.html
>
> Kitware offers ITK Training Courses, for more information visit:
> http://kitware.com/products/protraining.html
>
> Please keep messages on-topic and check the ITK FAQ at:
> http://www.itk.org/Wiki/ITK_FAQ
>
> Follow this link to subscribe/unsubscribe:
> http://www.itk.org/mailman/listinfo/insight-developers
>
>
> _______________________________________________
> Powered by www.kitware.com
>
> Visit other Kitware open-source projects at
> http://www.kitware.com/opensource/opensource.html
>
> Kitware offers ITK Training Courses, for more information visit:
> http://kitware.com/products/protraining.html
>
> Please keep messages on-topic and check the ITK FAQ at:
> http://www.itk.org/Wiki/ITK_FAQ
>
> Follow this link to subscribe/unsubscribe:
> http://www.itk.org/mailman/listinfo/insight-developers
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.itk.org/mailman/private/insight-developers/attachments/20101229/fa30d563/attachment.htm>