[Insight-users] about gaussian filter

[Luis Ibanez]

Hi Liu,

This filter is not implemented using convolutions.

Instead, it uses IIR filters in order to approximate
the effect of the convolution. This is the reason why
this filter is faster than traditional convolution.


You will find details about the family of Recursive Gaussian
filters:

http://www.itk.org/Insight/Doxygen/html/classitk_1_1RecursiveGaussianImageFilter.html

       in the original publication by Rachid Deriche:

"Recursively Implementing The Gaussian and Its Derivatives",
INRIA, 1993,
ftp://ftp.inria.fr/INRIA/tech-reports/RR/RR-1893.ps.gz


If you continue experiencing any problems with boundary effects,
then please post a screenshot of your output image in a public
web site or in one of the many image databases available out there
and let us know where to look for the image.  You could also post
a piece of source code that replicates the problem that you are
reporting.



     Regards,


        Luis


------------------
liu jianfei wrote:
> hi
> 
>    I used itkrecussivegaussiangradientmagnitude to compute the
> magnitude of a 3D volume. I tried to set the deviation as 5, 10, and
> 20. I found that the magnitude of the image border was very large even
> it's the background voxel. Since the image is first convolved with the
> gaussian filter, can someone tell me how many voxels are used to
> integrate the convolution. As far as I know, it's usually controlled
> by the decay of the gaussian function, for example, it could be one
> sigma (deviation), two, etc. Can someone tell me what's the time of
> sigma? Thanks
> 
> 
>    have a nice day!
> _______________________________________________
> Insight-users mailing list
> Insight-users at itk.org
> http://www.itk.org/mailman/listinfo/insight-users
> 
>