 |
 |
I am afraid not, but I can explain this here and hopefully one day it will make it to the wiki. When you resample an image (e.g. with FLIRT or APPLYWARP), there are several options, all of which more or less calculate a linear combination of scalars. In vecreg, you are taking linear combinations of vectors, so you have to be careful with the sign of vectors otherwise they cancel each other out. So what vecreg does is it turns the vectors into rank=1 matrices (x*x’), takes linear combinations of those, then takes the first eigenvector of the result (where the sign ambiguity arises). I hope this is clearer. Cheers Saad On 26 Oct 2016, at 16:36, Justin <[log in to unmask]<mailto:[log in to unmask]>> wrote: Saad, Thanks! Is there a paper or wiki article that discusses the interpolation used in vecreg? -Justin