[Insight-users] Smoothing Vector Images

[Sergio Vera]

Hello
Extracted from the itk doxygen documentation:

*As compared to
**itk::DiscreteGaussianImageFilter*<http://www.itk.org/Doxygen318/html/classitk_1_1DiscreteGaussianImageFilter.html>
*, this filter (**RecursiveGaussianImageFilter) **tends to be faster for
large kernels, and it can take the derivative of the blurred image in one
step. *

so if you are using smaller sigmas and hence, smaller Gaussians,
DiscreteGaussianImageFilter may result in faster execution time.

And now I'm speculating but my guess is that 1 second to perform
3x256x256x22 convolutions is not dramatically slow (it seems reasonable, as
those represents more than 4 millions of convolution operations), but i
don't know if it can be done much more faster.



On Wed, Jun 16, 2010 at 3:41 PM, Merrem, Andreas <
andreas.merrem at medma.uni-heidelberg.de> wrote:

> Hi Bill and Sergio,
>
> Thank you for your answers.
>
> I tried using nth element adaptors and the
> VectorIndexSelectionCastImageFilter, but it doesn't make it faster, what
> takes time is really the smoothing itself (as you already expected).
>
> Yes, I built both the itk library and the application in release mode. Do
> you think a second is a reasonable amount of time for smoothing 3 image
> volumes of size 256*256*22?
>
> If anyone knows a smoothing filter that does Gaussian smoothing and works
> faster than the itk recursive filter (perhaps something in frequency space)
> that would be a great help.
>
>
>
>
>
> -----Ursprüngliche Nachricht-----
> Von: Bill Lorensen [mailto:bill.lorensen at gmail.com]
> Gesendet: Dienstag, 15. Juni 2010 20:12
> An: Merrem, Andreas
> Cc: Luis Ibanez; insight-users at itk.org
> Betreff: Re: [Insight-users] Smoothing Vector Images
>
> First question: are you building ITK and your app with Release?
>
> On Tue, Jun 15, 2010 at 11:34 AM, Merrem, Andreas <
> andreas.merrem at medma.uni-heidelberg.de> wrote:
> > Hi Luis, hi everyone,
> >
> > I am trying to smooth 3D image volumes of size 256 * 256 * 22 pixels,
> > in which each pixel contains a 3 dimensional itk::CovariantVector. The
> > vector components should be smoothed independently over the whole
> > volume. My current solution is to first copy the 3 vector components
> > in each pixel into
> > 3 separate images with pixel type float, then use the
> > itk::RecursiveGaussianImageFilter for smoothing on all 3 images, and
> > finally copy the pixel values of the smoothed images back into the
> > covariant vector image.
> >
> > My problem is that this procedure takes about one second. As a part of
> > a registration program, this leads to extremely slow registration
> > (about 10 minutes for one 256*256*22 dicom volume). Is there a way to
> > use less memory by avoiding the copying, or do you know a different,
> > faster technique of applying a Gaussian filter?
> >
> > Thanks a lot,
> >
> > Andreas
> >
> > _____________________________________
> > 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://www.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-users
> >
> >
> _____________________________________
> 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://www.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-users
>



-- 
Sergio Vera

Alma IT Systems
C/ Vilana, 4B, 4º 1ª
08022 Barcelona
T. (+34) 932 380 592
www.alma3d.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.itk.org/pipermail/insight-users/attachments/20100616/acd2019b/attachment.htm>