[Insight-developers] Flawed logic in recent changes to SymmetricForcesDemonsRegistrationFunction

[Bill Lorensen]

Hans,

That was plenty quick!

Bill

On Fri, Nov 14, 2008 at 9:26 PM, Hans Johnson <hans-johnson at uiowa.edu> wrote:
> Sorry this took so long. My day got away from me.  The fix has been
> validated and committed.
>
> cvs commit -m"BUG: (0008098) Final change to needed fix orientation
> preserving code.  These changes fix an inproper refactoring."
> Code/Algorithms/itkSymmetricForcesDemonsRegistrationFunction.txx
> /cvsroot/Insight/Insight/Code/Algorithms/itkSymmetricForcesDemonsRegistratio
> nFunction.txx,v  <--
> Code/Algorithms/itkSymmetricForcesDemonsRegistrationFunction.txx
> new revision: 1.8; previous revision: 1.7
>
>
> Hans
>
> On 11/14/08 7:38 AM, "Bill Lorensen" <bill.lorensen at gmail.com> wrote:
>
>> Hans,
>>
>> First, thanks for spending the effort to get the registration
>> functions to work properly with direction cosines.
>>
>> I believe the changes you made to the subject class have some
>> problems. The TransformIndexToPhysicalPoint's code should not be
>> inside the "dim" for-loop. All "dim" elements of tmpIndex are not
>> computed when the TransformIndexToPhysicalPoint calls are made.
>>
>> I looked at this because of the failing tests:
>> http://www.cdash.org/CDash/testDetails.php?test=12475804&build=213744
>> and
>> http://www.cdash.org/CDash/testDetails.php?test=12476298&build=213744
>>
>> They really should not fail since their direction cosines are identity.
>>
>> Bill
>
>