[SciPy-user] Gaussian pyramid using ndimage

Alexander Borghgraef alexander.borghgraef.rma@gmail....
Fri Mar 23 04:38:48 CDT 2007

On 3/16/07, Zachary Pincus <zpincus@stanford.edu> wrote:
> I *think* that all of the filters in ndimage proceed by pre-filtering
> and then doing whatever task is required, so even if there was a
> downsampling filter (maybe 'zoom' would work), it wouldn't be memory-
> efficient in the way you were hoping.
> Also, the spline filters used by ndimage for pre-filtering seem
> somewhat broken:
> http://projects.scipy.org/scipy/scipy/ticket/213

 Hmm. So ndimage isn't really ready for use yet? Pity, image processing is
my main use for scipy.
Documentation should be extended too, it's not always clear what exactly an
algorithm is doing, what
the valid parameters are and what they're doing. I also noticed
generic_filter requires the filter function
to work on an unraveled array, and has to return a scalar, so you can't
create tensor images with it
(motionfields, gradient fields, in my case the 3D normal field for an
elevation map...). Some missing
functionality there, and definitely in need of better documentation.

Alex Borghgraef
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://projects.scipy.org/pipermail/scipy-user/attachments/20070323/2fb244ac/attachment.html 

More information about the SciPy-user mailing list