| contour fill [message #21378] |
Fri, 25 August 2000 00:00  |
so
Messages: 9
Registered: August 2000
|
Junior Member |
|
|
Hi,
I am trying to get IDL to isolate the values contained within a closed
contour of a plotted data array. Basically, I want to perform a
calculation using these values instead of cloouring them in like /fill
does.
Any help would be greatly appreciated!
Cheers, Stephen O'Sullivan
============================================================ ===============
Dr Stephen O'Sullivan | Telephone: +44-113-2335187
Department of Applied Mathematics | Facsimile: +44-113-2429925
University of Leeds | Email : so@amsta.leeds.ac.uk
Leeds LS2 9JT UK | WWW: www.amsta.leeds.ac.uk/applied
============================================================ ===============
|
|
|
|
|
|
|
|
|
|
| Re: contour fill [message #21448 is a reply to message #21378] |
Sat, 26 August 2000 00:00  |
wmc
Messages: 117
Registered: February 1995
|
Senior Member |
|
|
Stephen O'Sullivan <so@amsta.leeds.ac.uk> wrote:
> On Fri, 25 Aug 2000, David Fanning wrote:
>> Stephen O'Sullivan (so@amsta.leeds.ac.uk) writes:
>>> I am trying to get IDL to isolate the values contained within a closed
>>> contour of a plotted data array. Basically, I want to perform a
>>> calculation using these values instead of cloouring them in like /fill
>>> does.
> Thanks for your help but the problem that I didn't really make clear is
> that I need all the values contained within the specified closed contour,
> not just those above or below the contour value. In this sense it's a
> geometric thing.
Well.. wen you say *all* the values, you presumably mean te pixels at
a certain resolution, rather tan te infinite number of real points?
If so: wy not contour onto a pixmap, ten read in te pixmap (tvrd) ten
points_i_like=where(pixmap eq cosen_colour)
-W. (posting from a keyboard with a bad "h", you may have noticed...)
--
William M Connolley | wmc@bas.ac.uk | http://www.nerc-bas.ac.uk/icd/wmc/
Climate Modeller, British Antarctic Survey | Disclaimer: I speak for myself
|
|
|
|
|
|
| Re: contour fill [message #21462 is a reply to message #21378] |
Fri, 25 August 2000 00:00 |
Craig Markwardt
Messages: 1869
Registered: November 1996
|
Senior Member |
|
|
davidf@dfanning.com (David Fanning) writes:
> Stephen O'Sullivan (so@amsta.leeds.ac.uk) writes:
>
>> Thanks for your help but the problem that I didn't really make clear is
>> that I need all the values contained within the specified closed contour,
>> not just those above or below the contour value. In this sense it's a
>> geometric thing.
>
> Oh, well then, you are in for some work. :-)
>
> You will have to use the Path_Info and Path_XY keywords
> to fish out the contour lines you are concerned with
> ...
Wow David you must be trying hard to drum up consulting business :-).
I tried it myself, having never had ROI experience before, and got
something to work in about ten minutes.
Here is an example of thresholding an image the hard way, using the
technique you just described. The first part (the wordiest) is just
to set up an image to work on. The meat of the function is CONTOUR
and POLYFILLV -- and no graphics are produced. In fact you don't even
need to set up a graphics coordinate system on the screen. The
routines GAUSS2 and PLOTIMAGE are available from my web page
(http://cow.physics.wisc.edu/~craigm/idl/, check the Full Listing).
Once you have the list of pixels (the value of WH below) you can do
anything with it. I did a simple threshold but of course you could do
something harder. Now of course, the *real* problem is deciding
*which* contour to use when there are many of them. Here I have just
taken the first one. If you know there will always be one big one
then you might be able to get away by calculating the area of each
(eg, number of pixels in each) and taking the largest.
Craig
;; Create image using two gaussian functions
nx = 100 & ny = nx
dx = 0.1 & dy = dx
x = findgen(nx)*dx-5. & y = x
xx = rebin(reform(x,nx,1),nx,ny)
yy = rebin(reform(y,1,ny),nx,ny)
im = gauss2(xx, yy, [2.5, -4.2, 2., 50]) + gauss2(xx, yy, [-.5, 3, 4., 30])
;; Extract contour information - in this case everything inside the 20 contour
contour, im, x, y, levels=[20], path_info=pin, path_xy=pxy, $
/path_data, /close
;; Extract the contour of interest
c0off = pin(0).offset + lindgen( pin(0).n )
c0x = (pxy(0,c0off) - x(0)) / dx ;; Convert to scan cols & lines
c0y = (pxy(1,c0off) - y(0)) / dy
;; Use POLYFILLV to create a region of interest, and hence a mask
wh = polyfillv(c0x, c0y, nx, ny)
mask = (im*0)
mask(wh) = 1
;; Plot the masked image for verification
imgx = [-5.05, 4.95] & imgy = imgx
plotimage, im*mask, range=[0,52], imgxrange=imgx, imgyrange=imgy
Good luck,
Craig
--
------------------------------------------------------------ --------------
Craig B. Markwardt, Ph.D. EMAIL: craigmnet@cow.physics.wisc.edu
Astrophysics, IDL, Finance, Derivatives | Remove "net" for better response
------------------------------------------------------------ --------------
|
|
|
|
| Re: contour fill [message #21468 is a reply to message #21378] |
Fri, 25 August 2000 00:00 |
davidf
Messages: 2866
Registered: September 1996
|
Senior Member |
|
|
Stephen O'Sullivan (so@amsta.leeds.ac.uk) writes:
> Sounds like a nightmare!
I think you are beginning to get into the spirit
of IDL contours! :-)
> The reason I'm doing it is that I need some self-consistent way of
> defining a closed region in a fluid flow problem. It doesn't matter if
> it's ad hoc just so long as it's consistent.
For a closed contour, this is really not so bad. Er, well,
once you figure out how the Path_Info and Path_XY keywords
work. But PolyFillV is self-consistent if you take some
care in how you create your contour plot. For example,
always use the same size window, always position the contour
plot with the Position keyword, don't allow axis autoscaling,
create your own contour levels, etc. etc. All the normal
Contour things. :-)
Cheers,
David
--
David Fanning, Ph.D.
Fanning Software Consulting
Phone: 970-221-0438 E-Mail: davidf@dfanning.com
Coyote's Guide to IDL Programming: http://www.dfanning.com/
Toll-Free IDL Book Orders: 1-888-461-0155
|
|
|
|
|
|
|
|
|
|
| Re: contour fill [message #21533 is a reply to message #21378] |
Thu, 31 August 2000 03:46 |
so
Messages: 9
Registered: August 2000
|
Junior Member |
|
|
On 26 Aug 2000, Craig Markwardt wrote:
>
> davidf@dfanning.com (David Fanning) writes:
>> Leaving aside the observation that getting *something*
>> to work is not exactly the same as getting the *right*
>> thing to work, I'm still concerned that the poor fellow
>> is not working on the right thing. :-(
>>
>> Unless your goal is some kind of pseudo-science, working
>> with the representation of the data rather than the data
>> itself is rarely a good idea.
>
> He was asking *how?*. You are asking *why?*. Both are fair
> questions.
>
> You wondered why a basic threshold wouldn't work. I think his reply
> was that he wanted to select a *particular* contour. When there are
> several peaks in the data then a simple threshold will not work. As I
> pointed out the hardest part in the algorithm I gave is selecting the
> right contour. [And I gave a posible heuristic to decide.]
>
> Craig
>
> --
> ------------------------------------------------------------ --------------
> Craig B. Markwardt, Ph.D. EMAIL: craigmnet@cow.physics.wisc.edu
> Astrophysics, IDL, Finance, Derivatives | Remove "net" for better response
> ------------------------------------------------------------ --------------
>
>
That's right. I need a particular contour which is well defined and there
is a well rigorous physical meaning to what I am trying to do. It is
of course the representation of the data which I am manipulating and the I
fully understand that interpretation of this representation is important
if proper physical sense is to be made.
I haven't tried the suggestions which have been posted but when I have
I'll let you know how they work compared to a C program I have written to
do the job.
Cheers, Stephen
============================================================ ===============
Dr Stephen O'Sullivan | Telephone: +44-113-2335187
Department of Applied Mathematics | Facsimile: +44-113-2429925
University of Leeds | Email : so@amsta.leeds.ac.uk
Leeds LS2 9JT UK | WWW: www.amsta.leeds.ac.uk/applied
============================================================ ===============
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
