Hi,
There's no "right" answer.
It depends what you want and what you are doing.
Basically either case will give you very similar results.
Just think about the region from voxel 49 to 51 and
see what you think is more desirable.
All the best,
Mark
Maarten Versluis wrote:
> Thanks a lot that really helped. So one more question, if I want to
> rotate around the centre of the image, what is the correct way of
> calculating the offcentres?
> Let's say I have a 100 x 100 x 100 image with 1 mm voxels, is the
> translation to the centre then (50,50,50), or (50.5,50.5,50.5), or
> something else?
>
> Many thanks in advance!
>
> Maarten
>
> Maarten Versluis
>
>
> -----Original Message-----
> From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On
> Behalf Of Mark Jenkinson
> Sent: woensdag 19 november 2008 9:27
> To: [log in to unmask]
> Subject: Re: [FSL] Registration using only rotations, but no
> translations
>
> Hi Maarten,
>
> Yes - if you change the last column to be 0 0 0 1 then you are
> doing "no translation". However rotation and translation are
> highly coupled, so when you say "no translation" what you actually
> get is a rotation about the 0,0,0 voxel (which is in the corner of
> the image) and may not be what you want. If you want to do a
> rotation about a different centre, then you'll need to actually have
> a non-zero translation. The way to work out the translation is:
> trans = new_centre - rotation_matrix * old_centre
> where the rotation_matrix has the 0 0 0 1 final column.
>
> All the best,
> Mark
>
>
>
> On 19 Nov 2008, at 07:27, Maarten Versluis wrote:
>
>
>> Dear all,
>>
>> I am struggling with the following problem:
>>
>> I have 2 volumes of the same person scanned with a different
>> resolution. I want to register the low resolution image to the high
>> resolution image, using a rigid body 6 DOF model.
>> On a different dataset I would like apply only the found rotations
>> and skip the translations. What would be the correct way of doing
>> so? I was thinking of changing the last column of the rotation
>> matrix to [0 0 0 1], is that correct?
>> Many thanks in advance,
>>
>> Maarten
>> Maarten Versluis
>> Leiden University Medical Center
>> Department of Radiology - C3Q
>> Albinusdreef 2
>> 2333 ZA Leiden
>> The Netherlands
>> E-mail: [log in to unmask]
>>
>>
>
>
