The centre of rotation is the corner of the image - the one with
the (0,0,0) voxel coordinate. More specifically, it will be about
the centre of this voxel. If you want to see exactly what happens
to voxel coordinates (as reported by fslview) and flirt matrices
then you can use img2imgcoord.
All the best,
On 28 Oct 2008, at 22:47, Andre Gouws wrote:
> Hope you can clarify a detail in FLIRT for me.
> The short version: what is the centre-of-rotation used by flirt when
> applying a transform (4x4 .mat) to a volume?
> The full version: I have a surface generated from a structural MRI
> (call this Volume-A) using Freesurfer. I know the freesurfer surface's
> vertices have the same mm coordiante space as Volume-A.
> I want to get this surface into the coordinate frame of a different
> for the same participant (call it volume B). I use FLIRT to generate
> transform (.mat) describing the affine transformation of Volume-A to
> If I overlay the generated 'A_to_B Volume' on A the transformation is
> virtually perfect. If I however use the .mat coordinates to
> transform the
> vertices of my surface file, the fit is not as good. It seems to me
> this is because I am not applying the rotation values of the
> matrix to the correct centre-of-rotation in the coordinate frame of my
> freesurfer surface file.
> Hence the question: what is the centre-of-rotation used by flirt when
> applying a transform (4x4 .mat) to a volume? (end full version)
> Thanks in advance..