Dear Mark,
I have a closely related question: is a FLIRT affine matrix always
unique for 12 degrees of freedom, or does avscale have to make any
assumptions to recover the individuals rotations (etc.)? If there is a
definite solution, are the decomposition equations online anywhere?
(I've looked, but haven't found them...)
Thanks,
Jon
Mark Jenkinson wrote:
> Dear Jeff,
>
> Unfortunately this doesn't have an easy answer as it depends
> on certain conventions, such as where you want the centre
> of rotation to be. In fact, all the elements in the 3x3 submatrix
> are affected by rotations, and the centre of rotation plus the
> rotations themselves plus the desired translations affect the
> elements in the fourth column.
>
> If you want a decomposition using _our_ conventions, then
> run avscale --allparams on the matrix file and it will give you
> the rotations, translations (with respect to the centre of rotation
> being the lower left corner of the image) plus scales and skews.
>
> If you want to go the other way, you can create 6 dof matrices
> using the makerot utility.
>
> Hope this helps.
> All the best,
> Mark
>
>
>
> On 27 Sep 2006, at 19:25, Jeffrey Stanley wrote:
>
>> Dear FSLer's ... I need help in the definition of the terms in the 4x4
>> transformation matrix (a.k.a.,
>> the *.mat file) ... from what I can tell the M14, M24 and M34 elements
>> are the x, y,z translation
>> terms... M11, M22 and M33 elements are the respective scaling terms...
>> however, which elements
>> control the rotation and what are the unites... for example, what
>> would the elements be if I wanted
>> to rotate the image about the x-axis by 30 degrees???
>>
>> any help is greatly appreciated.
>>
>> Jeff
>>
|