[SciPy-User] Difference in quality from different interpolation orders

[Charles R Harris]

On Sat, Jun 18, 2011 at 10:40 AM, denis <denis-bz-gg at t-online.de> wrote:

> Folks,
>  a non-expert addon to David's and Zach's expert comments:
> splines can interpolate (go through the input data points exactly),
> or smooth but not interpolate exactly.
> Catmull-Rom splines interpolate, B-splines smooth more.
> One can mix the two, e.g. 2/3 C-R spline + 1/3 Bspline;
> there's a great paper describing this for cubic splines,
>    Mitchell and Netravali, "Reconstuction filters in computer
> graphics", 1988
>    http://portal.acm.org/citation.cfm?id=378514
>
> Does anyone know what splines ndimage.spline_filter uses --
> interpolating ?
>
>
Interpolating I'm pretty sure, as the images need to be prefiltered. I
suspect the smoothing splines can be got at by passing the unfiltered image
and prefilter=False to the routines, but I haven't tried it and it isn't
documented.

Chuck
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20110618/162830e1/attachment.html>