> The other issue is reslicing, since that's the most typical (pre)processing
> bottleneck. The SPM code itself looks good, but it appears to call some
> matlab routines. While these might be quite efficient algorithmically, I
> thought perhaps they could be speeded up by recoding them in C and then
> creating a mex file.
Most of the computations are actually done using spm_sample_vol and spm_slice_vol
mex files, which are very computationally expensive - especially with sinc
interpolation. These will eventually use B-spline interpolation, which takes a
fraction of the time with similar resulting image quality.
I don't know for certain how much of the time is involved in these mex files.
Maybe I should make use of the "profile" function in Matlab to get a bit more of
an idea.
Best regards,
-John
--
Dr John Ashburner.
Functional Imaging Lab., 12 Queen Square, London WC1N 3BG, UK.
tel: +44 (0)20 78337491 or +44 (0)20 78373611 x4381
fax: +44 (0)20 78131420 http://www.fil.ion.ucl.ac.uk/~john
|