Hi,
I'm glad the correlation is very strong as that means that it is using
the same centre and not moving it around in a data-dependent way.
There is no need to worry about internal parameters, and in fact it is more
difficult to save them. What is saved out to the file is the transformation
matrix, and this only has one type. Hence just stick to saving the
transformation matrix and then using this with applyxfm (it is the only
thing that can ever be used with applyxfm). These matrices will in general
show non-zero shift (though a different one from the parameters)
and you definitely want to keep these and not reset them. And they
are not artifactual, they are just about encoding the centre of the
operations in a coordinate system that has the origin of the coordinate
system elsewhere.
All the best,
Mark
On 24 Nov 2010, at 14:57, Silvia De Santis wrote:
> Thanks very very much Mark! In fact, Xshift turns out to be perfectly correlated to Yshear.
> But in this case, I'd really need to know if applyxfm wants the shift parameters expressed with respect to the lower left corner or to the centre of gravity. In other word, if I save the transformation parameters, shall I apply the transormation via applyxfm resetting to 0 Xshift or leaving the "artifactual" values?
>
> Thanks again and best regards,
> Silvia
|