[Insight-developers] [Fwd: [Fwd: Bug in Analyzer Writer]] : More on ImageDimension vs SpaceDimension

[Luis Ibanez]

Hi Rakesh,

Thanks a lot for your bug report.

We are actually discussing this in the Developers mailing list.

The issue is whether the Direction cosines of an image should
have dimensions NxN or not, and how they should be interpreted.

We had discussed so far the case of 2D images that are embedded
in a 3D space.

And interestingly, here you pointed out the opposite case, when
4D images are embedded in a 3D space.

That is, we can have direction cosines that are:


      Matrix<  ImageDimensions, SpaceDimensions >

for the case of 2D images embedded in 3D, the direction type
would be:

      Matrix<  2, 3 >

while for the case of 4D images embedded in a 3D space, the
direction type would be

      Matrix<  4, 3 >


The current proposal is to replace the trait of DirectionType
at the level of the ImageBase class, from

  Matrix<double, VImageDimension, VImageDimension> DirectionType;

to

  vnl_matrix< double >  DirectionType;


and at run time set the dimensions of the matrix to reflect
the combination of ImageDimension and SpaceDimension.


Please let us know if you would consider that this change
above may help to solve the problem you are observing.


    Thanks


       Luis


-------------- next part --------------
An embedded message was scrubbed...
From: Zack Galbreath <zack.galbreath at kitware.com>
Subject: [Fwd: Bug in Analyzer Writer]
Date: Tue, 23 Sep 2008 08:02:10 -0400
Size: 18014
URL: <http://www.itk.org/mailman/private/insight-developers/attachments/20080923/ad339127/attachment.eml>