[Insight-developers] Code / Common / itkScaleSkewVersor3DTransform.txx

[Stephen R. Aylward]

I'm backing out all changes.   The call to 3 instead of 2 was a typo. 
There are related tests and they appropriately failed last night.   The 
old implementation, however, is also wrong wrt not computing an explicit 
skew component matrix.

s

Simon Warfield wrote:
> 
> I don't think the changes to the itkScaleSkewVersor3DTransform.txx are 
> correct.
> Around line 228, m_Skew[3] has been initialized to 0.0, and it seems 
> intended for the user to manipulate only
> m_Skew[0], m_Skew[1] and m_Skew[2].  m_Skew[2] never appears in any 
> calculation.
> It appears to me that m_Skew[2] was intended where m_Skew[3] is written.
> 
> Furthermore, if this correction was made, I think the use of the three 
> parameters to provide the skew is incorrect.
> I don't think the skew as computed is what was intended.
> 
> 
> I also think it is wrong to utilize only 3 skew parameters since this 
> restricts the nature of transforms that can be constructed.
> I think it should be possible to construct an arbitrary affine transform 
> using this class.
> 
> 
> Changing the number of parameters in this transform breaks existing 
> code.  It is not backwards compatible.
> It appears that there is no test case or cases exercising the correct 
> operation of this transformation.
> 
> 
> I think the class should be called
> itkShearScaleVersor3DTransform
>  and that six Shear parameters should be provided.
> 
> 

-- 
=============================================================
Stephen R. Aylward, Ph.D.
Chief Medical Scientist
Kitware, Inc. - Chapel Hill Office
http://www.kitware.com
Phone: (518)371-3971 x300