| Re: linear-space image converted to log-space [message #58835] |
Sat, 23 February 2008 20:46 |
jkj
Messages: 48
Registered: April 2007
|
Member |
|
|
> model_2->Add,o2
> model_2->Scale, 1, 100, 1 ;<<<< cheat here to make effect big enough
> to see
to make them both the same height:
scale = ny / alog10(ny)
model_2->Scale, 1, scale, 1
|
|
|
|
| Re: linear-space image converted to log-space [message #58836 is a reply to message #58835] |
Sat, 23 February 2008 19:33  |
jkj
Messages: 48
Registered: April 2007
|
Member |
|
|
On Feb 23, 8:26 pm, "ben.bighair" <ben.bigh...@gmail.com> wrote:
> On Feb 23, 5:45 pm, jkj <ke...@vexona.com> wrote:
>
>
>
>> Hi,
>
>> I have a need to reproduce an image like this:
>
>> http://diamondhead.org/linear_logspace.png
>
>> beginning with data like this:
>
>> http://diamondhead.org/linear_linearspace.png
>
>> Using logarithmic scaling, the first image (produced by a custom
>> graphics package) expands lower frequency components while contracting
>> the higher frequency components. To do this in object graphics, I
>> believe I will have to reconstruct the image data so that the lower
>> frequency [the bottom] information is replicated according to a log
>> scaling scheme.
>
>> I can think of some [tortured] ways to do this but keep thinking there
>> must be some sophisticated way to handle this in IDL. Any thoughts
>> would be appreciated.
>
>> Thanks,
>> -Kevin
>
> Hi,
>
> I think I would use a flat filled IDLgrSurface that is projected as
> one might view an image - with a log scaled y axis. Then use the
> TEXTURE_MAP property to show the image data.
>
> The example below uses XOBJVIEW to show the same image twice. The
> bottom is "regular" and the top is stretched like in your example.
> Note that I had to stretch the second model so you can see the
> effect. I can't recall at this late hour how to get better control of
> that, but it can be done much better.
>
> Ben
>
> **BEGIN
> rose = READ_IMAGE(FILEPATH('rose.jpg',
> SUBDIRECTORY=['examples','data']))
> dim = SIZE(rose,/DIM)
> nx = dim[1] & ny = dim[2]
> x = findgen(nx)+1
> y = findgen(ny)+1
> s = replicate(1,nx,ny)
>
> ;the first surface is "regular"
> o1 = OBJ_NEW("IDLgrSurface", s, x, y, $
> color = [255,255,255], style = 2, $
> texture_map = obj_new("IDLgrImage", rose))
> model_1 = OBJ_NEW("IDLgrModel")
> model_1->Add,o1
> model_1->Translate, 0, -ny/2. - 10, 0
>
> ;the second surface is log scale in y
> o2 = OBJ_NEW("IDLgrSurface", s, x, alog10(y), $
> color = [255,255,255], style = 2, $
> texture_map = obj_new("IDLgrImage", rose))
> model_2 = OBJ_NEW("IDLgrModel")
> model_2->Add,o2
> model_2->Scale, 1, 100, 1 ;<<<< cheat here to make effect big enough
> to see
> model_2->Translate, 0,ny/2. + 10, 0
>
> xobjview, [model_1, model_2] , /BLOCK
>
> OBJ_DESTROY, [model_1, model_2]
> **END
My hero! :-)
Very gracious of you - thanks!
-Kevin
|
|
|
|
| Re: linear-space image converted to log-space [message #58838 is a reply to message #58836] |
Sat, 23 February 2008 18:26 |
ben.bighair
Messages: 221
Registered: April 2007
|
Senior Member |
|
|
On Feb 23, 5:45 pm, jkj <ke...@vexona.com> wrote:
> Hi,
>
> I have a need to reproduce an image like this:
>
> http://diamondhead.org/linear_logspace.png
>
> beginning with data like this:
>
> http://diamondhead.org/linear_linearspace.png
>
> Using logarithmic scaling, the first image (produced by a custom
> graphics package) expands lower frequency components while contracting
> the higher frequency components. To do this in object graphics, I
> believe I will have to reconstruct the image data so that the lower
> frequency [the bottom] information is replicated according to a log
> scaling scheme.
>
> I can think of some [tortured] ways to do this but keep thinking there
> must be some sophisticated way to handle this in IDL. Any thoughts
> would be appreciated.
>
> Thanks,
> -Kevin
Hi,
I think I would use a flat filled IDLgrSurface that is projected as
one might view an image - with a log scaled y axis. Then use the
TEXTURE_MAP property to show the image data.
The example below uses XOBJVIEW to show the same image twice. The
bottom is "regular" and the top is stretched like in your example.
Note that I had to stretch the second model so you can see the
effect. I can't recall at this late hour how to get better control of
that, but it can be done much better.
Ben
**BEGIN
rose = READ_IMAGE(FILEPATH('rose.jpg',
SUBDIRECTORY=['examples','data']))
dim = SIZE(rose,/DIM)
nx = dim[1] & ny = dim[2]
x = findgen(nx)+1
y = findgen(ny)+1
s = replicate(1,nx,ny)
;the first surface is "regular"
o1 = OBJ_NEW("IDLgrSurface", s, x, y, $
color = [255,255,255], style = 2, $
texture_map = obj_new("IDLgrImage", rose))
model_1 = OBJ_NEW("IDLgrModel")
model_1->Add,o1
model_1->Translate, 0, -ny/2. - 10, 0
;the second surface is log scale in y
o2 = OBJ_NEW("IDLgrSurface", s, x, alog10(y), $
color = [255,255,255], style = 2, $
texture_map = obj_new("IDLgrImage", rose))
model_2 = OBJ_NEW("IDLgrModel")
model_2->Add,o2
model_2->Scale, 1, 100, 1 ;<<<< cheat here to make effect big enough
to see
model_2->Translate, 0,ny/2. + 10, 0
xobjview, [model_1, model_2] , /BLOCK
OBJ_DESTROY, [model_1, model_2]
**END
|
|
|
|
| Re: linear-space image converted to log-space [message #58839 is a reply to message #58838] |
Sat, 23 February 2008 18:04 |
jkj
Messages: 48
Registered: April 2007
|
Member |
|
|
>
> Is it not just image = alog10(image), with the proper use of finite()
> and where() to trap errors?
No [but thanks!], that would take the data into log space and leave
the same number of pixels representing a given "row/frequency" of the
data. I need to expand the low frequency portion of the image so that
portions which only received 1 pixel now are smeared out over multiple
pixels in the low freq. range. The data itself does not change, just
the percentage of the image it receives - low frequency elements are
smeared out broadly while high frequency elements (less interesting in
this case) are highly compressed.
I suspect I should be able to do this with interpolate, so that is
what I'm trying to pencil out.
|
|
|
|
| Re: linear-space image converted to log-space [message #58840 is a reply to message #58839] |
Sat, 23 February 2008 15:32 |
Vince Hradil
Messages: 574
Registered: December 1999
|
Senior Member |
|
|
On Feb 23, 4:45 pm, jkj <ke...@vexona.com> wrote:
> Hi,
>
> I have a need to reproduce an image like this:
>
> http://diamondhead.org/linear_logspace.png
>
> beginning with data like this:
>
> http://diamondhead.org/linear_linearspace.png
>
> Using logarithmic scaling, the first image (produced by a custom
> graphics package) expands lower frequency components while contracting
> the higher frequency components. To do this in object graphics, I
> believe I will have to reconstruct the image data so that the lower
> frequency [the bottom] information is replicated according to a log
> scaling scheme.
>
> I can think of some [tortured] ways to do this but keep thinking there
> must be some sophisticated way to handle this in IDL. Any thoughts
> would be appreciated.
>
> Thanks,
> -Kevin
Is it not just image = alog10(image), with the proper use of finite()
and where() to trap errors?
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
