This very ancient spatial normalisation procedure is based on doing a least
squares fitting. Because one image may be scaled very differently from
another, the procedure includes an additional intensity scaling parameter to
account for this.
Affine registration is done by Gauss-Newton, although there may be a small
amount of Levenberg-Marquardt regularisation involved.
Best regards,
-John
On Tuesday 14 October 2008 09:24, Raouia Ayachi wrote:
> Hi SPMers,
>
> I am reading the paper of Friston 95 and Ashburner 97, and I don't
> understand why in the spatial normalization step, an additional parameter
> is added to estimate intensity mapping between the images. Could you please
> explain to me this process?
>
>
> Another point please, in fact I am using SPM5 and in the manual I didn't
> find which optimization problem have been used in the affine
> transformation. Is it The Gauss Newton one?
>
> Thanks for your help experts,
>
> Best regards.
>
>
> Raouia AYACHI
|