| Smooth and /EDGE_TRUNCATE [message #88457] |
Tue, 29 April 2014 07:35  |
Fabzi
Messages: 305
Registered: July 2010
|
Senior Member |
|
|
IDLers,
I've lost 3 hours of my life on this. Can someone explain me why the
/EDGE_TRUNCATE keyword has an incidence on the results WITHIN the image,
where the kernel meets no edge?
pro test_smooth
; make an array with a nan in the center
array = FINDGEN(5,5)
array[2,2] = !VALUES.F_NAN
print, ' Exp 1'
print, 'Expected', MEAN(array[1:3,1:3], /NAN)
print, 'No truncate', (smooth(array, 3, /NAN))[2,2]
print, 'Truncate', (smooth(array, 3, /NAN, /EDGE_TRUNCATE))[2,2]
; so far so good. Add a NaN somewhere else
array[1,1] = !VALUES.F_NAN
print, ' Exp 2'
print, 'Expected', MEAN(array[1:3,1:3], /NAN)
print, 'No truncate', (smooth(array, 3, /NAN))[2,2]
print, 'Truncate', (smooth(array, 3, /NAN, /EDGE_TRUNCATE))[2,2]
end
Thanks a lot!
Fabien
|
|
|
|
| Re: Smooth and /EDGE_TRUNCATE [message #88459 is a reply to message #88457] |
Tue, 29 April 2014 09:04  |
David Fanning
Messages: 11724
Registered: August 2001
|
Senior Member |
|
|
Fabien writes:
> I've lost 3 hours of my life on this. Can someone explain me why the
> /EDGE_TRUNCATE keyword has an incidence on the results WITHIN the image,
> where the kernel meets no edge?
>
> pro test_smooth
>
> ; make an array with a nan in the center
> array = FINDGEN(5,5)
> array[2,2] = !VALUES.F_NAN
>
> print, ' Exp 1'
> print, 'Expected', MEAN(array[1:3,1:3], /NAN)
> print, 'No truncate', (smooth(array, 3, /NAN))[2,2]
> print, 'Truncate', (smooth(array, 3, /NAN, /EDGE_TRUNCATE))[2,2]
>
> ; so far so good. Add a NaN somewhere else
> array[1,1] = !VALUES.F_NAN
> print, ' Exp 2'
> print, 'Expected', MEAN(array[1:3,1:3], /NAN)
> print, 'No truncate', (smooth(array, 3, /NAN))[2,2]
> print, 'Truncate', (smooth(array, 3, /NAN, /EDGE_TRUNCATE))[2,2]
>
> end
Not sure this "explains" it, but this section of the documentation might
have at least saved you a couple of hours. :^)
Note: Normally, two-dimensional floating-point arrays are smoothed in
one pass. If any of the EDGE_* keywords are specified for a two-
dimensional floating-point array, the result is obtained in two passes,
first for all of the rows, and second for all of the columns. Therefore,
the results for points in the interior of the array may differ slightly
when any of the EDGE_* keywords are set. This difference will be most
pronounced if the array contains NaN values.
Cheers,
David
--
David Fanning, Ph.D.
Fanning Software Consulting, Inc.
Coyote's Guide to IDL Programming: http://www.idlcoyote.com/
Sepore ma de ni thue. ("Perhaps thou speakest truth.")
|
|
|
|
| Re: Smooth and /EDGE_TRUNCATE [message #88461 is a reply to message #88459] |
Tue, 29 April 2014 09:23 |
Fabzi
Messages: 305
Registered: July 2010
|
Senior Member |
|
|
On 29.04.2014 18:04, David Fanning wrote:
> Not sure this "explains" it, but this section of the documentation might
> have at least saved you a couple of hours. :^)
OMG, "obvious" as allways. My opinion surely has absolutely no value,
but this means to me that the EDGE_* keywords are more than useless:
they are also dangerous.
Note, also, that CONVOL doesn't have this issue:
print, 'Convol', (convol(array, fltarr(3,3)+1./9, /NAN, $
/EDGE_TRUNCATE, /NORMALIZE))[2,2]
Produces the correct results.
Not for the future: "smooth" is ok for using on your last holiday's
picures, not for real data filtering ;)
Thanks dave!
Fabien
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
