[Insight-users] VersorRigid3DTransform

[Luis Ibanez]

Hi Radhika,

The VersorRigid3DTransformOptimizer has alredy
been commited.  However its test is currently
failing so it may not be ready for use.

You can check on the state of the test by looking
at the Dashboard. Just click on the "Tests" button
and search for "VersorRigid3DTransformOptimizer".

Today the test for this class looks like:
http://www.itk.org/Testing/Dashboard/20040105-0500-Nightly/TestDetail/__Testing_Code_Numerics_itkVersorRigid3DTransformOptimizerTest.html

You can also add your email address to the CC fields
of the bug entry in the bug tracker. In this way you
will be notified whenever changes are logged in the
bug database concerning this issue.


For using a sequence of transforms:


1) If you start with a TranslationTransform,
    the RegularStepGradientDescentOptimizer is
    probably the best choice.

2) For pure rotation you could go with the
    VersorTransform and its associated optimizer
    the VersorTransformOptimizer. However, note
    that this transform assumes the origin to be
    the center of rotation.

3) You are right, you should resample at the
    end of each intermediate step, and you will
    loose some information in the process of
    resampling. What you may want to do is to
    accept these looses while you are searching
    for the intermediate components of the transform
    (translation, rotation... etc) and once you
    find them, you start from scratch by incorporating
    these intermediate transforms as the initialization
    of an fully AffineTransform. In that way, your final
    refinement is using the original information of the
    moving image.



Regards,


   Luis


---------------------
Radhika Sivaramakrishna wrote:

> Hi Luis,
> 
> Thanks for your detailed responses on the optimizer for the 
> VersorRigid3DTransform.
> 
>  
> 
> Could you let me know the timeframe when you anticipate this optimizer 
> to be available. Is there a way to check for it on my own for eg. on the 
> dashboard?
> 
>  
> 
>  
> 
> I also had a related question. In the software guide, you recommend that 
> when there are multiple transformations to take care of, it is sometimes 
> better
> 
> to take care of them one at a time using independent transforms in order 
> to reduce the dimension of the space of optimization.
> 
> So in my case, if I wanted to take care of translation first, followed 
> by the rotation part
> 
> after that, my questions are:
> 
>  
> 
> 1)       For the translation part would it be better to use the 
> RegularStepGradientDescentOptimizer?
> 
> 2)       For the pure rotation part, would it better to use the 
> VersorTransformOptimizer?
> 
> 3)       After I finish the translation part, I am assuming that I will 
> need to do a resampling stage so I can create an intermediate
> 
>       image, which I then use as my moving image for the rotation part. 
> Is this correct? If so, I would totally need two resamplings. Would I 
> not lose some detail in
> 
>         my final image because of that. Is there a better way to do 
> this, so I can compose the two transformations, so I only do one resampling?
> 
>  
> 
> Thanks
> 
> Radhika
> 
>  
> 
> -----------------------------------------------------
> 
> Confidentiality Notice.
> 
> This email message is for the sole use of the intended recipient(s) and 
> may contain confidential and privileged information. Any unauthorized 
> review, use, disclosure or distribution is prohibited. If you are not 
> the intended recipient, please contact the sender by reply email and 
> destroy all copies of the original message. If you are the intended 
> recipient, please be advised that the content of this message is subject 
> to access, review and disclosure by the sender's Email System Administrator.
>