AW: [Insight-users] GradientAnisotropicDiffusionImageFilter

Joshua Cates cates at sci.utah.edu
Tue Jun 29 12:54:49 EDT 2004


Hi,

The edge-preserving smoothing filters all differ from one another.  The
results depend on the data and the parameters.  See the itkSoftwareGuide
for some additional information and references.  The differences may be 
subtle in many cases.

If you are interested in speed, you might try the itkBilateralImageFilter.  
Among the pde filters, GradientAnisotropicDiffusion filter will be
slightly faster than CurvatureAnisotropicDiffusion, and you can run it at
double the time step (.125 in 3D versus .06).  You can safely ignore the
warning as long as you don't exceed a time step of .125.


Josh.

On Tue, 29 Jun 2004, salah wrote:

> Thanks Joshua!
> 
> Yes, the problem was probably that my image has USHORT pixel
> type. I did something like 
> 
> image ===> cast to float ===> edge-pre-smooth ===> cast back to UShort ===> smoothed image
> 
> and it worked. However, the warning about the time step still appears, 
> but the result is correct.
> 
> A question please:
> Do the different edge-preserving smoothing filters differ much 
> from each other? I mean from the result point of view? Which one 
> is known to be the fastest?
> 
> Many thanks,
> 
> Zein
> 
> > -----Ursprüngliche Nachricht-----
> > Von: Joshua Cates [mailto:cates at sci.utah.edu]
> > Gesendet: Monday, June 28, 2004 6:01 PM
> > An: salah
> > Cc: ITK Users (E-Mail)
> > Betreff: Re: [Insight-users] GradientAnisotropicDiffusionImageFilter
> > 
> > 
> > Hi Zein,
> > 
> > A time step of 0.125 should be fine with the Gradient... filter and 3D
> > data.  (The CurvatureAnisotropicDiffusion... filter is a 
> > different story.)
> > This warning is probably being generated in error.  I'll have a look.
> > 
> > I can't explain why you would get no change in the output.  
> > Rest assured 
> > these filters are tested nightly and should be working fine.  
> > Note that 
> > the conductance and number of iterations required for a 
> > reasonable result 
> > will vary widely with the data. 
> > 
> > I suggest you experiment with a small chunk of your data (so 
> > that you can 
> > do iterations quickly) to get appropriate settings for your data.
> > 
> > One more thing: be sure to convert your data to floating point values 
> > before filter.  This is a pde filter and the algorithm 
> > doesn't make sense 
> > with integer types.
> > 
> > hope this helps,
> > 
> > Josh.
> > 
> > 
> > On Mon, 28 Jun 2004, salah wrote:
> > 
> > > Hello all,
> > > 
> > > I am trying to perform edge-preserving smoothing by the 
> > > 
> > > GradientAnisotropicDiffusionImageFilter
> > > 
> > > I used the following parameter values:
> > > 
> > > filter->SetNumberOfIterations( 5 );
> > > filter->SetTimeStep( 0.125 );
> > > filter->SetConductanceParameter( 3 );
> > > 
> > > 1. I got a warning that 0.125 is not a stable value and 
> > that the minimum
> > >    stable values is 0.0625.
> > > 
> > > 2. However with both these values and the other parameter 
> > values, the 
> > >    output image is just the same as the input image. I 
> > tested this with
> > >    CT input images of USHORT and UCHAR pixel types.
> > > 
> > > What could the problem be ? what are appropriate values to 
> > use with this
> > >    Filter? Does the other edge-preserving smoothing filters differ?
> > > 
> > > 
> > > Many thanks,
> > > 
> > > Zein
> > > 
> > > _______________________________________________
> > > Insight-users mailing list
> > > Insight-users at itk.org
> > > http://www.itk.org/mailman/listinfo/insight-users
> > > 
> > 
> > 
> 



More information about the Insight-users mailing list