| Pain with the contour() function [message #94497] |
Wed, 14 June 2017 08:09  |
Helder Marchetto
Messages: 520
Registered: November 2011
|
Senior Member |
|
|
Hi,
I was trying to plot an image with the contour function and I banged my head against the monitor for a while. I now have the solution and I'm sharing it. Probably most people know this very well. I didn't.
I followed the example given for the colorbar() function (Example: Discrete Contour Levels with Colorbar):
http://www.harrisgeospatial.com/docs/Colorbars.html
So I generated my data with the code below and it appeared strangely shifted to the side.
dis = dist(688)
n_levels = 6
levels = findgen(n_levels)
ct_number = 4
ct_indices = bytscl(levels)
loadct, ct_number, rgb_table=ct, /silent
step_ct = congrid(ct[ct_indices, *], 256, 3)
dis = (n_levels-1)*dis/max(dis)
ii = contour(dis, c_value = levels, rgb_table = step_ct, rgb_indices = ct_indices, /fill, axis_style=0)
I then started playing around with the position and margin keywords, but had no luck. Finally it all comes down to using xRange and yRange (or xStyle=1, yStyle=1):
ii = contour(dis, c_value = levels, rgb_table = step_ct, rgb_indices = ct_indices, /fill, axis_style=0, xStyle=1, yStyle=1)
The reason is that contour() plots images as if they were plots, so it defines some axis around it and you have to make sure you're not having uncovered regions.
Well, back to work.
Cheers,
Helder
|
|
|
|
| Re: Pain with the contour() function [message #94498 is a reply to message #94497] |
Wed, 14 June 2017 08:45  |
Markus Schmassmann
Messages: 129
Registered: April 2016
|
Senior Member |
|
|
On 06/14/2017 05:09 PM, Helder wrote:
> I was trying to plot an image with the contour function and I banged my head against the monitor for a while. I now have the solution and I'm sharing it. Probably most people know this very well. I didn't.
> I followed the example given for the colorbar() function (Example: Discrete Contour Levels with Colorbar):
> http://www.harrisgeospatial.com/docs/Colorbars.html
>
> So I generated my data with the code below and it appeared strangely shifted to the side.
>
> dis = dist(688)
> n_levels = 6
> levels = findgen(n_levels)
> ct_number = 4
> ct_indices = bytscl(levels)
> loadct, ct_number, rgb_table=ct, /silent
> step_ct = congrid(ct[ct_indices, *], 256, 3)
> dis = (n_levels-1)*dis/max(dis)
> ii = contour(dis, c_value = levels, rgb_table = step_ct, rgb_indices = ct_indices, /fill, axis_style=0)
>
> I then started playing around with the position and margin keywords, but had no luck. Finally it all comes down to using xRange and yRange (or xStyle=1, yStyle=1):
>
> ii = contour(dis, c_value = levels, rgb_table = step_ct, rgb_indices = ct_indices, /fill, axis_style=0, xStyle=1, yStyle=1)
>
> The reason is that contour() plots images as if they were plots, so it defines some axis around it and you have to make sure you're not having uncovered regions.
>
> Well, back to work.
another pain with contour, or any combination of raster graphics and
vector graphics elements is, that they are natively offset by half a
pixel (or whatever you want to call the data unit here):
c=contour(dist(10),overplot=image(35*dist(10),dimension=[250 ,250], $
position=[25,25,225,225],/dev))
the solution to that is
contour, dist(10), path_xy=line, path_info=info, /path_data_coord, $
closed=0,levels=[0:6]
i1=image(35*dist(10),dimension=[250,250], $
position=[25,25,225,225],/dev)
foreach in,info do p=plot(line[*,in.offset+[lindgen(in.n),in.type ? 0 :$
!null]]+.5, color=255b-[40b,40b,0b]*byte(in.value),overplot=i1)
-- Markus
|
|
|
|
| Re: Pain with the contour() function [message #94507 is a reply to message #94498] |
Thu, 15 June 2017 01:56 |
lecacheux.alain
Messages: 325
Registered: January 2008
|
Senior Member |
|
|
Le mercredi 14 juin 2017 17:45:39 UTC+2, Markus Schmassmann a écrit :
> On 06/14/2017 05:09 PM, Helder wrote:
>> I was trying to plot an image with the contour function and I banged my head against the monitor for a while. I now have the solution and I'm sharing it. Probably most people know this very well. I didn't.
>> I followed the example given for the colorbar() function (Example: Discrete Contour Levels with Colorbar):
>> http://www.harrisgeospatial.com/docs/Colorbars.html
>>
>> So I generated my data with the code below and it appeared strangely shifted to the side.
>>
>> dis = dist(688)
>> n_levels = 6
>> levels = findgen(n_levels)
>> ct_number = 4
>> ct_indices = bytscl(levels)
>> loadct, ct_number, rgb_table=ct, /silent
>> step_ct = congrid(ct[ct_indices, *], 256, 3)
>> dis = (n_levels-1)*dis/max(dis)
>> ii = contour(dis, c_value = levels, rgb_table = step_ct, rgb_indices = ct_indices, /fill, axis_style=0)
>>
>> I then started playing around with the position and margin keywords, but had no luck. Finally it all comes down to using xRange and yRange (or xStyle=1, yStyle=1):
>>
>> ii = contour(dis, c_value = levels, rgb_table = step_ct, rgb_indices = ct_indices, /fill, axis_style=0, xStyle=1, yStyle=1)
>>
>> The reason is that contour() plots images as if they were plots, so it defines some axis around it and you have to make sure you're not having uncovered regions.
>>
>> Well, back to work.
>
> another pain with contour, or any combination of raster graphics and
> vector graphics elements is, that they are natively offset by half a
> pixel (or whatever you want to call the data unit here):
>
> c=contour(dist(10),overplot=image(35*dist(10),dimension=[250 ,250], $
> position=[25,25,225,225],/dev))
>
> the solution to that is
>
> contour, dist(10), path_xy=line, path_info=info, /path_data_coord, $
> closed=0,levels=[0:6]
> i1=image(35*dist(10),dimension=[250,250], $
> position=[25,25,225,225],/dev)
> foreach in,info do p=plot(line[*,in.offset+[lindgen(in.n),in.type ? 0 :$
> !null]]+.5, color=255b-[40b,40b,0b]*byte(in.value),overplot=i1)
>
> -- Markus
I would rather write:
c=contour(dist(10),overplot=image(35*dist(10),dimension=[250 ,250], $
IMAGE_LOCATION=[-0.5,-0.5], $
position=[25,25,225,225],/dev))
since, by convention, the image grid locates the lower left point of each pixel.
alx.
|
|
|
|
| Re: Pain with the contour() function [message #94508 is a reply to message #94498] |
Thu, 15 June 2017 01:57 |
Helder Marchetto
Messages: 520
Registered: November 2011
|
Senior Member |
|
|
On Wednesday, June 14, 2017 at 5:45:39 PM UTC+2, Markus Schmassmann wrote:
> On 06/14/2017 05:09 PM, Helder wrote:
>> I was trying to plot an image with the contour function and I banged my head against the monitor for a while. I now have the solution and I'm sharing it. Probably most people know this very well. I didn't.
>> I followed the example given for the colorbar() function (Example: Discrete Contour Levels with Colorbar):
>> http://www.harrisgeospatial.com/docs/Colorbars.html
>>
>> So I generated my data with the code below and it appeared strangely shifted to the side.
>>
>> dis = dist(688)
>> n_levels = 6
>> levels = findgen(n_levels)
>> ct_number = 4
>> ct_indices = bytscl(levels)
>> loadct, ct_number, rgb_table=ct, /silent
>> step_ct = congrid(ct[ct_indices, *], 256, 3)
>> dis = (n_levels-1)*dis/max(dis)
>> ii = contour(dis, c_value = levels, rgb_table = step_ct, rgb_indices = ct_indices, /fill, axis_style=0)
>>
>> I then started playing around with the position and margin keywords, but had no luck. Finally it all comes down to using xRange and yRange (or xStyle=1, yStyle=1):
>>
>> ii = contour(dis, c_value = levels, rgb_table = step_ct, rgb_indices = ct_indices, /fill, axis_style=0, xStyle=1, yStyle=1)
>>
>> The reason is that contour() plots images as if they were plots, so it defines some axis around it and you have to make sure you're not having uncovered regions.
>>
>> Well, back to work.
>
> another pain with contour, or any combination of raster graphics and
> vector graphics elements is, that they are natively offset by half a
> pixel (or whatever you want to call the data unit here):
>
> c=contour(dist(10),overplot=image(35*dist(10),dimension=[250 ,250], $
> position=[25,25,225,225],/dev))
>
> the solution to that is
>
> contour, dist(10), path_xy=line, path_info=info, /path_data_coord, $
> closed=0,levels=[0:6]
> i1=image(35*dist(10),dimension=[250,250], $
> position=[25,25,225,225],/dev)
> foreach in,info do p=plot(line[*,in.offset+[lindgen(in.n),in.type ? 0 :$
> !null]]+.5, color=255b-[40b,40b,0b]*byte(in.value),overplot=i1)
>
> -- Markus
Hi,
I see your problem. This seems to have been discussed here:
https://groups.google.com/d/msg/comp.lang.idl-pvwave/8mTxEIM evzE/RIplXmvm9N4J
Your solution works, but at the cost of losing the contour function managing all the contour lines at once.
Helder
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
