[Insight-users] Re: QuaternionRigidTransform

[Junyi Xia]

>Hi All,
>
>I have a question about choosing a transform type for my CT-MR image
>registration. I have got the voxel size information of the sample datasets
>(from DICOM header), do I need to use similarity transform, or only rescale
>them to the same voxel size before registration to use rigid transform.
>  
>
You don't need to resample them to the same voxel size because the 
registration works in the physical coordinates. Pls refer to the 
software guide  chapter 8.3 for more information

>What is the difference between QuaternionRigidTransform and
>Similarity3DTransform, I think QuaternionRigidTransform should also be
>able to address homogeneous scaling problem. BTW,  Is it necessary to use
>affine transform for multimodal registration?
>  
>
Whether using affine or rigid transform depends on the image dataset. If 
there is scaling or skewing between two dataset, affine transform is 
necessary. If there is certain deformation between the two, you need to 
consider deformable registration. You could try the rigid transform 
first, and see how well the registration goes.

>Thanks in advance.
>
>Yan
>
>
>  
>