On Monday 15 December 2003 12:24, Bertram Walter wrote:
> Dear SPM people,
>
> after reading of two papers by Thevenaz et al I have two questions about
> spline interpolation:
>
> - Our preprocessing steps include realignment without reslicing and
> b-spline interpolation after normalisation as recommended by John
> Ashburner. Thevenaz favours in his comparison of interpolation methods
> clearly b-spline for several reasons. But he restricts his discussion to
> data "regularely sampled on a Cartesian grid". After the normalisation
> (before resampling) the images are warped by DCTs and not regularly
> sampled. Q: Is it justified to use b-spline interpolation on warped data?
The original data are on a regular grid. Resampling involves interpolating
these values at non-integer co-ordinates.
Ideally, the code should probably have some kind of spatially varying
low-pass filtering before resampling in order to reduce aliasing effects.
This is probably only really needed if the warped images are at a much
lower resolution than the original ones.
>
> - Discussing image artifacts Thevenaz reports ringing artifacts especially
> with high-quality interpolation because their synthesis functions are
> oscillating. How does the degree of b-splines affects such a ringing?
Higher degree interpolation gives more ringing, but I don't think that this
is necessarily a bad thing - particularly as the data are usually smoothed
afterwards. It could be bad if negative values were truncated (e.g. if the
image was stored as unsigned), as this would bias the background of images
with lots of ringing to be higher intensity.
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
|