[Insight-users] ConvolutionImageFilter only works with scalar images?

[David Doria]

> This
> http://www.insight-journal.org/browse/publication/774
> or this
> http://www.itk.org/Doxygen/html/classitk_1_1NthElementImageAdaptor.html
>
> would work coupled with
> http://www.itk.org/Doxygen/html/classitk_1_1Function_1_1Compose3DCovariantVector.
> html
>
> Re: the ConvolutionImageFilter:  There are some other issues with it too that
> could be fixed up.  E.g., the internal use of 'double' in
> itkConvolutionImageFilter.txx.  Also, itkImageKernelOperator could be reworked --
> if the kernel image does not have the size implied by the operator radius it
> causes a crash.
>
> Regards,
> Matt

@Robert

ComposeRGBImageFilter is exactly the idea, but the Software Guide
warns against using RGBPixel if you want to treat each channel
separately, so I have always heeded that warning.

@Matt,

Perfect!

I put examples of this on the wiki:
Disassembling (extracting a component):
http://www.itk.org/Wiki/ITK/Examples/ImageProcessing/NthElementImageAdaptor

Assembling (combining components):
http://www.itk.org/Wiki/ITK/Examples/ImageProcessing/Compose3DCovariantVectorImageFilter

Should you add a bug to Mantis for the other issues you mentioned? It
would still be nice not have to do this disassembling and assembling
manually - would the things you mentioned allow this to work with
N-component images?

Thanks for the help,

David