[Scipy-tickets] [SciPy] #213: ndimage.rotate creates artifacts (general problem in ndimage.spline_filter)

SciPy scipy-tickets@scipy....
Fri Mar 23 01:40:05 CDT 2007


#213: ndimage.rotate creates artifacts (general problem in ndimage.spline_filter)
---------------------------+------------------------------------------------
 Reporter:  stefan         |        Owner:  somebody
     Type:  defect         |       Status:  new     
 Priority:  normal         |    Milestone:          
Component:  scipy.ndimage  |      Version:  devel   
 Severity:  normal         |   Resolution:          
 Keywords:                 |  
---------------------------+------------------------------------------------
Comment (by zachrahan):

 This ticket should be probably be closed, as I think that it reflects no
 more than a mis-understanding of the pre-filter step on my part.
 Essentially, transforms of images with "sharp edges", like the synthetic
 examples attached, should be done with spline interpolation of order 0 or
 1; higher-order splines will necessarily introduce ringing into these sort
 of images.

 I was incorrect regarding the spline pre-filter being at fault because I
 assumed that that pre-filter behaved like a traditional anti-aliasing pre-
 filter (i.e. that it should blur the image as a band-limiting step); the
 fact that it did not blur the images (even the smooth natural image in the
 second attachment) I took as evidence of the problem. However, (I think)
 the spline "pre-filter" really is computing the spline coefficients, and
 not "pre-filtering" the image in a traditional sense -- leading to my
 mistake.

 The final comment in this ticket (splines using mirror boundary
 conditions) should probably be re-added as a new feature request ticket
 (if desired).

-- 
Ticket URL: <http://projects.scipy.org/scipy/scipy/ticket/213#comment:3>
SciPy <http://www.scipy.org/>
SciPy is open-source software for mathematics, science, and engineering.


More information about the Scipy-tickets mailing list