[Insight-users] Re: A problem of registration

Luis Ibanez luis.ibanez at kitware.com
Sat Sep 11 15:08:27 EDT 2004


Hi Cao Jie,


   Transforms are not smart,
         they are not supposed to be.     :-)


The problem that you are confronting is related to the initialization
of the Center of Rotation. It seems that you are not initializing the
center of rotation of the transform and therefore the origin is being
used as the default center of rotation.

Please read the section 8.5 "Centered Transforms" of the SoftwareGuide

      http://www.itk.org/ItkSoftwareGuide.pdf

pdf-page 263-281. This will help you to understand the issues related
to the selection of the center of rotation when performing registration.
In particular you may want to use the CenteredTransformInitializer class
in order to initialize the center of rotation of the VersorRigid3D
Transform.

You will find several examples on how to do this in the directory

               Insight/Examples/Registration

The coding examples are described in detail in the ITK Software Guide.


  Regards,


     Luis




-------------
曹杰 wrote:

> Hi,
>    Thank you for your help recently. Now I have a serious problem that I can't solve:
>    When I used the VersorRigid3DTransform ,Mattes MutualInformation Metric
> and VersorRigid3DTransformOptimizer to registrate.
>    The tempImage is 256*256*33 (MRI0001~MRI0030) and the fixedImage is 
> 256*256*30 (MRI0004~MRI0030).So the fixedImage is the bottom part of the movingImage. Then the tempImage  rotates around the Z-axis by 0.1 degree
> and take a little translation in the X-Y plane---(0,5,-5)translation and becomes the movingImage.
>    The origins of fixedImage and movingImage are the same----(0,0,0).
>    I set the initial transform to (0,0,0,0,0,0) and the rotation center to
> (0,0,0).
>    The correct registration result should have a translation in the Z direction. However,the translation in the result is nearly (0,0,0).
> I think the translation is compensated by the rotation around the X-axis and
> Y-axis. Obviously,the result is wrong,at least bad.
>    Is it the problem that VersorRigid3DTransform isn't so smart and can't solve this kind of registration?
>    Thank you!
>                                                                   Cao Jie
>    
> 
> -------------------------------------------------------------------------------------
> AOC(冠捷)17寸液晶16ms¥2888 http://design.163.com/aocmonitor/index.html






More information about the Insight-users mailing list