[Insight-developers] Contrib: itkAverageImageFilter w/ example and tests

[Luis Ibanez]

Hi Torsten,

Thanks for contributing the AverageImageFilter.

We are moving into implementing the Insight Journal
as the entry gate for new classes into ITK.

The idea is that even simple classes will be posted
as papers accompanied by code and data.

When the classes are relatively simple, then the
papers will be short. For example, a couple of pages
that describe the rationale behind the need for a
filter and some guidelines on how it was implemented.

Once new classes pass the automatic testing of the
Insight Journal, and get positive reviews from the
community, then they will be moved into the toolkit.

We are trying to find a balance on this process that
will make easy for contributors to send their algorithms
and code. That is, we don't want this process to become
too bureaucratic. It is simply a way of increasing the
coherence of the Toolkit.

----

About your question regarding Traits, we have not
encountered the need for a "rounding" conversion
between RealType and other types.  This seems to be
too specific to the filters for being put in the
numeric traits.

What most filters do is to explicitly invoke rounding
or truncation when necessary, using vnl_math methods.


    Regards,


       Luis



=======================
Torsten Rohlfing wrote:
> 
> Hi --
> 
> Here's something I couldn't find in ITK so far and therefore implemented 
> myself: a filter to average an arbitrary number of images (all on the 
> same grid). Class sources, example command line application, and test 
> (with patch file for test integration) are in
> 
>  http://www.stanford.edu/~rohlfing/itk_averageimagefilter.tar.gz
> 
> The filter should handle a large range of pixel types, including vector 
> images etc. (see class documentation for a list of necessary pixel type 
> operators). It would be nice if someone could put this code into the CVS 
> repository sometimes, even if it takes until after the coming release.
> 
> As a note on the side - it seems that the itk::NumericTraits do not 
> provide a way to determine whether a real value should be rounded or not 
> when assigned to any given type. Is there something for that purpose 
> elsewhere, or has it just not been necessary so far?
> 
> Best,
>  Torsten
>