| Re: ROT is ROTTEN [message #28079] |
Tue, 20 November 2001 22:32 |
Wayne Landsman
Messages: 117
Registered: January 1997
|
Senior Member |
|
|
>
> ROT is bad. Can it be fixed? Is there a (fast) alternative?
The easiest alternative is to use the intrinsic ROTATE function, which is
specifically designed to deal with rotations of multiples of 90 degress.
A more subtle alternative is to add the MISSING keyword when using the ROT()
function, e.g.
print, total(rot(a,270,/interp, MISSING = !VALUES.F_NAN))
and you will find that there are values flagged as "missing". This is
because unless the rotation is exactly a multiple of 90 degrees, then there
will be subpixels in the output array for which there are no corresponding
values to interpolate in the input array. Instead of extrapolating,
these pixels get flagged as "missing" and the values returned by ROT() are
not to be trusted. (Exactly which pixels get flagged as missing depends
on the roundoff error.)
This question has come up before here -- perhaps RSI should modfiy ROT() so
that it calls ROTATE when the user supplies an exact integer mulitple of 90
degrees.
Wayne Landsman landsman@mpb.gsfc.nasa.gov
P.S. I would also like a keyword to ROT() to control the output dimensions
e.g. ftp://idlastro.gsfc.nasa.gov/landsman/idl/rot.pro
|
|
|
|
| Re: ROT is ROTTEN [message #28081 is a reply to message #28079] |
Tue, 20 November 2001 22:06  |
Richard French
Messages: 173
Registered: December 2000
|
Senior Member |
|
|
Bhautik Jitendra Joshi wrote:
>
> MOO>a=findgen(5,5) & print, a
> 0.00000 1.00000 2.00000 3.00000 4.00000
> 5.00000 6.00000 7.00000 8.00000 9.00000
> 10.0000 11.0000 12.0000 13.0000 14.0000
> 15.0000 16.0000 17.0000 18.0000 19.0000
> 20.0000 21.0000 22.0000 23.0000 24.0000
>
> MOO>print, total(a)
> 300.000
> MOO>print, total(rot(a,360))
> 262.000
>
Well, this is interesting! In IDL5.5, I get 262.000 on Windows98 and
Tru64.
In IDL5.4 on Windows, I get 247.000, but 262.000 on Tru64.
Anyone with earlier versions want to weigh in
here? It is hard to see why a rotation of 360 would not give you
what you started with, unless I am really missing something
fundamental...
(I just compared the source code on the 5.4 and 5.5 lib versions of rot
and they are identical except in the comment fields...)
Now if you do:
IDL> print,total(rot(dindgen(5,5),360.d0))
242.00000
IDL> print,!version
{ alpha OSF unix Compaq Tru64 5.5 Aug 28 2001 64 64}
IDL> print,total(rot(dindgen(5,5),360.d0))
282.00000
IDL> print,!version
{ x86 Win32 Windows Microsoft Windows 5.5 Aug 28 2001 32 64}
and here is a nice one... if you put the angle in integer degrees, you
get
one answer, in double precsission, you get another...
IDL> print,total(rot(dindgen(5,5),360))
262.00000
IDL> print,total(rot(dindgen(5,5),360.d0))
242.00000
IDL> print,!version
{ alpha OSF unix 5.4 Sep 25 2000 64 64}
I modified rot.pro to do things in double precision but it did not make
any difference. I agree that when you turn on /cubic you don't get
300.00, but something close to it. I have not displayed the images
to see what they look like, but it surely is not doing what we think
it should!
Dick
French
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
