[Insight-developers] Gaussian filters

[Joshua Cates]

Hi Mark,

I will add more documentation on DiscreteGaussianImageFilter as soon as I
have some spare minutes.  In the meantime, ignore the MaximumError
parameter.  The default setting is fine for most applications.

Currently, the DiscreteGaussianImageFilter will only give reasonable
output for real numbers (floats, doubles) so you need to use
CastImageFilter to convert your input to floats (and back).  In the
future this conversion will be handled internally within the filter
and be transparent to the user.  I will be sure to include this caveat in
the documentation.

Thanks for the feedback,

Josh.

______________________________
 Josh Cates
 School of Computer Science
 University of Utah
 Email: cates@sci.utah.edu
 Phone: (801) 587-7697
 URL:   www.cs.utk.edu/~cates


On Wed, 24 Jul 2002, mark foskey wrote:

> I'm new to the group and still learning ITK, and I've noticed a couple
> of apparent anomalies with the various Gaussian filter classes.
>
> 1. RecursiveGaussianImageFilter has a SetSigma method, while
> DiscreteGaussianImageFilter has a SetVariance method.  I think it would
> be helpful if the method were the same in both filters.
>
> 2. DiscreteGaussianImageFilter yielded an apparently blank image for all
> variance values I tried other than 0.  I am including the test program I
> used below.  I may be doing something wrong.
>
> 3. Perhaps most serious is that it was hard to figure out how to use
> these filters from their documentation.  Programmers are likely to
> assume that RecursiveGaussianImageFilter simply blurs an image; only if
> they read the documentation for RecursiveSeparableImageFilter will they
> learn that the filter must be applied once for each dimension.  Also,
> for DiscreteGaussianImageFilter, the SetMaximumError method isn't
> documented.  I'm not clear on what the m_MaximumError parameter controls
> in this context.  I have not read the code yet, but the point is that I
> think the meaning of the parameter should be documented.
>
> 4. It would be nice if there were a convenience filter that would
> internally chain together instances of the RecursiveGaussianImageFilter
> for each dimension, since most users will want to filter along all n dims.
>
> I'm sorry for my first message to the group to come across as a
> complaint, but I'm seeing ITK somewhat as a new user will, and I think
> we have to be aware things that frustrate beginners.
>
> Here's the test program:
>
> #include <itkImage.h>
> #include <itkImageFileReader.h>
> #include <itkImageFileWriter.h>
> #include <itkPNGImageIO.h>
> #include <itkDiscreteGaussianImageFilter.h>
>
> typedef itk::Image<unsigned char, 2> ImageType ;
>
> int main(int argc, char* argv[])
> {
>       if (argc < 2)
>       {
>           std::cout
>               << "usage: VerySimpleFilter [inputfile] [outputfile]"
>               << std::endl;
>           exit(0);
>       }
>
>       // Create an IO object for PNG files, to be used by both the
>       // reader and the writer.
>       typedef itk::PNGImageIO PNGIOType;
>       PNGIOType::Pointer png = PNGIOType::New();
>
>       // Instantiate a reader for the file.
>       typedef itk::ImageFileReader<ImageType> ImageReaderType;
>       ImageReaderType::Pointer reader = ImageReaderType::New();
>       reader->SetImageIO(png);
>       reader->SetFileName(argv[1]);
>
>       // Instantiate a filter and attach it to the reader.
>       typedef itk::DiscreteGaussianImageFilter<ImageType, ImageType>
>           GaussianFilterType;
>       GaussianFilterType::Pointer myFilter = GaussianFilterType::New();
>       myFilter->SetInput(reader->GetOutput());
>       myFilter->SetVariance(0.1);
>
>       // Instantiate a writer and attach it to the filter.
>       typedef itk::ImageFileWriter<ImageType> ImageWriterType;
>       ImageWriterType::Pointer writer = ImageWriterType::New();
>       writer->SetImageIO(png);
>       writer->SetInput(myFilter->GetOutput());
>       writer->SetFileName(argv[2]);
>
>       writer->Write();
>
>       return 0;
>
> }
> --
> Mark Foskey    (919) 843-5436   Computer-Aided Diagnosis and Display Lab
> mark_foskey@unc.edu             Department of Radiology, CB 7515, UNC
> http://www.cs.unc.edu/~foskey   Chapel Hill, NC  27599-7515
>
>
> _______________________________________________
> Insight-developers mailing list
> Insight-developers@public.kitware.com
> http://public.kitware.com/mailman/listinfo/insight-developers
>