| Map projection of IMAGE() is behaving strangely... [message #86512] |
Fri, 15 November 2013 05:37  |
andeh
Messages: 23
Registered: April 2011
|
Junior Member |
|
|
Hello,
First time poster, long time reader.
Using IDL 8.2 on linux, I have recently been experimenting with the new IDL IMAGE() function and discovered strange behaviour when projecting an image using the MAP_PROJECTION property (given in the final plot):
;; Set up variables
z = BINDGEN(3,5,6) * 21b
lon = FINDGEN(5)*20-60
lat = FINDGEN(6)*20-60
;; Just plot the image.
im0 = IMAGE( z, $
AXIS_STYLE=2, $
TITLE='Basic image', $
LAYOUT=[2,2,1] )
;; Now add coordinates for bottom left point in each pixel.
im1 = IMAGE( z, lon, lat, $
AXIS_STYLE=2, $
TITLE='Add x/y coordinates', $
LAYOUT=[2,2,2], /CURRENT )
;; Let IMAGE() know that we are using lat-lon coordinates.
im2 = IMAGE( z, lon, lat, $
AXIS_STYLE=2, $
GRID_UNITS=2, $
TITLE='+ GRID_UNITS=2', $
LAYOUT=[2,2,3], /CURRENT )
;; And add a map projection (Sinusoidal in this case, but the same
;; result is seen for Geographic, Orthographic, Interrupted Goode
;; and any other projection I've tried).
im3 = IMAGE( z, lon, lat, $
AXIS_STYLE=2, $
GRID_UNITS=2, $
MAP_PROJECTION='Sinusoidal', $
TITLE='+ MAP_PROJECTION = Sinusoidal', $
LAYOUT=[2,2,4], /CURRENT )
The longitude (x) coordinate appears to be mis-aligned so that the image pixels are stretched in that direction and the image area is filled by z[*,1:-1,*] instead of the full image z[*,*,*]. This doesn't happen in the latitude direction.
Am I missing something fundamental about the MAP_PROJECTION property?
Cheers,
Andy
|
|
|
|
| Re: Map projection of IMAGE() is behaving strangely... [message #86516 is a reply to message #86512] |
Fri, 15 November 2013 07:06  |
David Fanning
Messages: 11724
Registered: August 2001
|
Senior Member |
|
|
Andrew Smith writes:
> The longitude (x) coordinate appears to be mis-aligned so that the image pixels are stretched in that direction and the image area is filled by z[*,1:-1,*] instead of the full image z[*,*,*]. This doesn't happen in the latitude direction.
>
> Am I missing something fundamental about the MAP_PROJECTION property?
It seems to me the weirdness is not from the map projection, which I
think might actually be doing the *right* thing. I think the problem is
with the Image function. In this call:
;; Now add coordinates for bottom left point in each pixel.
im1 = IMAGE( z, lon, lat, $
AXIS_STYLE=2, $
TITLE='Add x/y coordinates', $
LAYOUT=[2,2,2], /CURRENT )
The image display shows coordinates that go from -60 to 60 in latitude,
and from -60 to 40 in longitude, but this isn't what is in the lat/lon
variables:
IDL> minmax, lat
MinMax: -60.0000 40.0000
IDL> minmax, lon
MinMax: -60.0000 20.0000
So, when you get to the map projection, you are getting the wrong
picture of what is suppose to be happening.
Don't know exactly. I just know something *always* seems to be weird
with function graphics routines when you look closely at them. ;-)
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: Map projection of IMAGE() is behaving strangely... [message #86517 is a reply to message #86516] |
Fri, 15 November 2013 07:37 |
andeh
Messages: 23
Registered: April 2011
|
Junior Member |
|
|
On Friday, 15 November 2013 15:06:11 UTC, David Fanning wrote:
> The image display shows coordinates that go from -60 to 60 in latitude,
> and from -60 to 40 in longitude, but this isn't what is in the lat/lon
> variables:
>
> IDL> minmax, lat
> MinMax: -60.0000 40.0000
> IDL> minmax, lon
> MinMax: -60.0000 20.0000
>
Hi David,
Thanks for your response. I guess this means that I am incorrect that the lat/lon values represent the bottom-left point in each pixel once we get to the map projection. I expected the upper range of the boxes to be 20 degrees larger than the maximum values as that was my bin width in the image.
Just to check, my output looks like this: http://goo.gl/1aY3ab
Did yours?
Moving the first column of data to the left of 60W line seems like a very strange feature of the projection... Especially since the xrange and all aspects of the y plotting behave as I was expecting.
Cheers,
Andy
|
|
|
|
|
|
| Re: Map projection of IMAGE() is behaving strangely... [message #86519 is a reply to message #86517] |
Fri, 15 November 2013 07:52 |
David Fanning
Messages: 11724
Registered: August 2001
|
Senior Member |
|
|
AJAS writes:
> Thanks for your response. I guess this means that I am incorrect that the lat/lon values represent the bottom-left point in each pixel once we get to the map projection. I expected the upper range of the boxes to be 20 degrees larger than the maximum values as that was my bin width in the image.
Well, hard to say without more investigation. I'm never particularly
optimistic when it comes to IDL map projections, though. They always
seem to be doing something different from what I *think* they should be
doing. :-)
> Just to check, my output looks like this: http://goo.gl/1aY3ab
> Did yours?
More or less, yes, but I'm still not absolutely convinced the problem is
in the map projection part of the code. In any case, have to run. One of
the kids has strep and has no "soft" food in the house. Running to the
rescue now. :-)
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: Map projection of IMAGE() is behaving strangely... [message #86525 is a reply to message #86512] |
Fri, 15 November 2013 08:49 |
MP
Messages: 15
Registered: March 2006
|
Junior Member |
|
|
On Friday, November 15, 2013 6:37:52 AM UTC-7, AJAS wrote:
>
> ;; And add a map projection (Sinusoidal in this case, but the same
> ;; result is seen for Geographic, Orthographic, Interrupted Goode
> ;; and any other projection I've tried).
> im3 = IMAGE( z, lon, lat, $
> AXIS_STYLE=2, $
> GRID_UNITS=2, $
> MAP_PROJECTION='Sinusoidal', $
> TITLE='+ MAP_PROJECTION = Sinusoidal', $
> LAYOUT=[2,2,4], /CURRENT )
>
This is an aside, and may not be useful in this case, but I strongly recommend using MAP, then IMAGE with /OVERPLOT. Although they should behave identically, I've had more success with this technique than using the MAP_PROJECTION keyword to IMAGE.
mp
|
|
|
|
| Re: Map projection of IMAGE() is behaving strangely... [message #86526 is a reply to message #86518] |
Fri, 15 November 2013 08:52 |
brainwasher
Messages: 1
Registered: November 2013
|
Junior Member |
|
|
On Friday, 15 November 2013 16:43:33 UTC+1, AJAS wrote:
> Woops!
>
>
>
> I just noticed the bottom Y value is missing as well! Disaster.
Well, that problem in not only related to IMAGE function. It is found also in iTools and function iImage, and it is inherited from who knows where and when. I was really expecting that new graphics will solve it, but when i tried it for the first time I was dissapointed.
I was always using a quite straightforward way around, to give to idl one row and column it can swallow:
z_enlarged = FLTARR(3,nlon+1, nlat+1)
z_enlarged[0,1,1] = z
lon_enlarged = ((max(lon)-min(lon))/float(nlon))*findgen(nlon+1)+min(lon)
lat_enlarged = ((max(lat)-min(lat))/float(nlat))*findgen(nlat+1)+min(lat)
It is not perfect still, because idl shrinks the field little bit, but with a large number of points it looks good.
Any other solution?
|
|
|
|
|
|
| Re: Map projection of IMAGE() is behaving strangely... [message #86531 is a reply to message #86526] |
Fri, 15 November 2013 09:43 |
Andy Sayer
Messages: 127
Registered: February 2009
|
Senior Member |
|
|
Brain, I wonder if that could be related at all to the odd behaviour (reported to ExelisVis but not something they're going to fix) of map_image truncating edges? https://groups.google.com/forum/#!searchin/comp.lang.idl-pvw ave/map_image/comp.lang.idl-pvwave/3BmVLThB8Lk/_QUaZ--wehkJ
On Friday, November 15, 2013 11:52:07 AM UTC-5, brain...@gmail.com wrote:
> On Friday, 15 November 2013 16:43:33 UTC+1, AJAS wrote:
>
>> Woops!
>
>>
>
>>
>
>>
>
>> I just noticed the bottom Y value is missing as well! Disaster.
>
>
>
> Well, that problem in not only related to IMAGE function. It is found also in iTools and function iImage, and it is inherited from who knows where and when. I was really expecting that new graphics will solve it, but when i tried it for the first time I was dissapointed.
>
>
>
> I was always using a quite straightforward way around, to give to idl one row and column it can swallow:
>
>
>
> z_enlarged = FLTARR(3,nlon+1, nlat+1)
>
> z_enlarged[0,1,1] = z
>
> lon_enlarged = ((max(lon)-min(lon))/float(nlon))*findgen(nlon+1)+min(lon)
>
> lat_enlarged = ((max(lat)-min(lat))/float(nlat))*findgen(nlat+1)+min(lat)
>
>
>
> It is not perfect still, because idl shrinks the field little bit, but with a large number of points it looks good.
>
>
>
> Any other solution?
|
|
|
|
| Re: Map projection of IMAGE() is behaving strangely... [message #94796 is a reply to message #86531] |
Tue, 17 October 2017 11:33 |
Brian McNoldy
Messages: 35
Registered: July 2000
|
Member |
|
|
I see this thread is nearly 4 years old now, but it just came in handy for me this week. I was trying to plot an image with a map projection and then overlay filled continents... all using functional graphics. It was frustrating, but I ended up using a combination of two tricks suggested in here:
1) I needed to add a blank row to the beginning of my image array because the map projection process would "eat"/truncate the first row (but not column). Doing so magically solved that quirk. No idea what's going on there.
2) I called MAP first with my projection of choice, and then added the IMAGE with /OVERPLOT.
The guts of my successful result looks like:
m=map('Mollweide',center_longitude=clon,limit=limit,position =position,$
dimensions=[1200,500])
i=image(array,lons[0:-2],lats,grid_units='degrees',rgb_table =table,$
image_dimensions=[360,90],image_location=[0,-45],/overplot)
(My data array spanned all longitudes but only -45 to +45 in latitude.)
(The different treatment of the lons and lats arrays was because of needing to extend the array and latitude vector as described in (1)).
It feels like a hack, but the result is what I was hoping for.
FYI,
Brian
On Friday, November 15, 2013 at 12:43:52 PM UTC-5, AMS wrote:
> Brain, I wonder if that could be related at all to the odd behaviour (reported to ExelisVis but not something they're going to fix) of map_image truncating edges? https://groups.google.com/forum/#!searchin/comp.lang.idl-pvw ave/map_image/comp.lang.idl-pvwave/3BmVLThB8Lk/_QUaZ--wehkJ
>
> On Friday, November 15, 2013 11:52:07 AM UTC-5, brain...@gmail.com wrote:
>> On Friday, 15 November 2013 16:43:33 UTC+1, AJAS wrote:
>>
>>> Woops!
>>
>>>
>>
>>>
>>
>>>
>>
>>> I just noticed the bottom Y value is missing as well! Disaster.
>>
>>
>>
>> Well, that problem in not only related to IMAGE function. It is found also in iTools and function iImage, and it is inherited from who knows where and when. I was really expecting that new graphics will solve it, but when i tried it for the first time I was dissapointed.
>>
>>
>>
>> I was always using a quite straightforward way around, to give to idl one row and column it can swallow:
>>
>>
>>
>> z_enlarged = FLTARR(3,nlon+1, nlat+1)
>>
>> z_enlarged[0,1,1] = z
>>
>> lon_enlarged = ((max(lon)-min(lon))/float(nlon))*findgen(nlon+1)+min(lon)
>>
>> lat_enlarged = ((max(lat)-min(lat))/float(nlat))*findgen(nlat+1)+min(lat)
>>
>>
>>
>> It is not perfect still, because idl shrinks the field little bit, but with a large number of points it looks good.
>>
>>
>>
>> Any other solution?
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
