On Friday, August 9, 2002, at 05:32 PM, Erik-Jan Vlieger wrote:
>> I would also question whether there would be *no* movement at all.
>> Some small movement (say 0.5mm) could easily occur, in which case it
>> would
>> be better to run some registration - wouldn't it?
>
> You are absolutely right. On the other hand, the minimization algorithm
> would
> start much closer to the minumum, I expect the speed of the registration
> could improve. Would you expect the same?
Yes, using this as an initial estimate and then bypassing the global search
phase (using -nosearch) will definitely speed it up a bit. The question is
whether it is worth all the hassle of figuring out where to save these
parameters
or just let the registration work it out for you.
Mind you, for 2D slices from a 3D volume there is no choice - you must have
the scanner parameters to get a reasonable fit. There might be other cases
like this too, so maybe it is worth it.
Best of luck,
Mark
|