[Insight-developers] image filters versus functions

[Luis Ibanez]

Hi Leila,

The Functions are intended to be used when you need to evaluate
a sparse collection of voxels, and therefore using a filter that
computes the full image results in computing a lot of pixels that
are not going to be used.

You may want to measure the trade-off as:

1) P : percentage of pixels in the image for which you actually
        need to know the Gaussian blur values

2) Tb : Time that a Gaussian filter will take to blur your image

3) Tv : Time that the Gaussian function takes for one pixel

4) N  : Number of pixels in the image.



    Then compare

             Tb    versus   Tv * N * P


Usually functions are a better deal when you are doing things
like vessel tracking, since you evaluate a very small fraction
of the pixels in the image.


Note also that different Gaussian filters have different behaviors.
For example, some of them depend on the value of Sigma.



Please let us know what you find.


    Thanks


      Luis


---------------------
Leila Baghdadi wrote:
> Hi everyone,
> 
> I was just trying itkGaussianBlurImageFunction on an image and I noticed
> it takes a long time since you have to Evaluate the function at every
> voxel.
> 
> I am just wondering if there is a way to speed up the function or
> possibly use an image filter instead.
> 
> Thanks
> 
> Leila
> 
> _______________________________________________
> Insight-developers mailing list
> Insight-developers at itk.org
> http://www.itk.org/mailman/listinfo/insight-developers
>