| Re: find_boundary routine [message #38636] |
Thu, 18 March 2004 18:47  |
David Fanning
Messages: 11724
Registered: August 2001
|
Senior Member |
|
|
Brian Frizzelle writes:
> Let me apologize if this post is a bit lengthy. And in case it's
> important, I'm using IDL 6.0 on a Windows machine.
>
> I need help calculating the perimeters of regions. I have an input
> array of roughly 900 by 1200 cells which contains 15 different
> categories (coded 1 through 15). For each category, I need to find the
> total perimeter length among all of its regions.
>
> I found David Fanning's FIND_BOUNDARY routine and ran it on this
> array, but it seemed to stop at the first region that it found. So I
> wrote a script that subsetted each array first by category, then by
> region within the category (regions were created with LABEL_REGION).
> Running FIND_BOUNDARY again, I got an output for each category, but
> the numbers seemed off.
>
> So I resorted to creating a simple 5x5 array with two categories, one
> region each, to test the perimeter output of FIND_BOUNDARY.
> - Category 1 has 16 cells and a perimeter of 18.
> - Category 2 has 9 cells and a perimeter of 18.
> * FIND_BOUNDARY output a perimeter of 2 for each category.
> * FIND_BOUNDARY also only give me an output array with 2 X/Y
> coordinates.
>
> Has anyone else found this to be a problem? Maybe I'm passing in the
> wrong parameters. Here are the details of what I've done so far. Any
> help would be greatly appreciated.
Here is an example of a 5 by 5 array, with the middle 9 pixels
making up the region:
IDL> array=[[0,0,0,0,0],[0,1,1,1,0],[0,1,1,1,0],[0,1,1,1,0],[0,0, 0,0,0]]
IDL>indices = Where(array GT 0)
IDL>pts = Find_Boundary(indices, XSize=5, YSize=5, Perimeter=p)
IDL>Print, p
8.0
I had to make a small change to the Find_Boundary to accommodate this
few a number of indices. I had to change line 316 from this:
boundaryPts = IntArr(2, Long(xsize) * ysize / 4L)
To this:
boundaryPts = IntArr(2, Long(xsize) * ysize / 2L)
You might think the correct "perimeter" would be 9 in this
case, but you have to remember that the algorithm is calculating
from the center of each pixel. A quick sketch of these 9 pixels
and their centers will convince you that the correct answer is 8.
Other combinations of pixels in my tests gave similar correct
results. I think the algorithm is working properly.
Cheers,
David
--
David W. Fanning, Ph.D.
Fanning Software Consulting, Inc.
Coyote's Guide to IDL Programming: http:/www.dfanning.com/
Phone: 970-221-0438, IDL Book Orders: 1-888-461-0155
|
|
|
|
| Re: find_boundary routine [message #38642 is a reply to message #38636] |
Thu, 18 March 2004 14:47  |
btt
Messages: 345
Registered: December 2000
|
Senior Member |
|
|
Brian Frizzelle wrote:
> Let me apologize if this post is a bit lengthy. And in case it's
> important, I'm using IDL 6.0 on a Windows machine.
>
> I need help calculating the perimeters of regions. I have an input
> array of roughly 900 by 1200 cells which contains 15 different
> categories (coded 1 through 15). For each category, I need to find the
> total perimeter length among all of its regions.
>
> I found David Fanning's FIND_BOUNDARY routine and ran it on this
> array, but it seemed to stop at the first region that it found. So I
> wrote a script that subsetted each array first by category, then by
> region within the category (regions were created with LABEL_REGION).
> Running FIND_BOUNDARY again, I got an output for each category, but
> the numbers seemed off.
>
> So I resorted to creating a simple 5x5 array with two categories, one
> region each, to test the perimeter output of FIND_BOUNDARY.
> - Category 1 has 16 cells and a perimeter of 18.
> - Category 2 has 9 cells and a perimeter of 18.
> * FIND_BOUNDARY output a perimeter of 2 for each category.
> * FIND_BOUNDARY also only give me an output array with 2 X/Y
> coordinates.
>
> Has anyone else found this to be a problem? Maybe I'm passing in the
> wrong parameters. Here are the details of what I've done so far. Any
> help would be greatly appreciated.
>
> Brian Frizzelle
>
> -------------------------
>
> arrayTest: regCat1: regCat2:
> 1 1 1 2 2 1 1 1 0 0 0 0 0 1 1
> 1 1 1 2 2 1 1 1 0 0 0 0 0 1 1
> 1 1 1 1 2 1 1 1 1 0 0 0 0 0 1
> 1 1 1 1 2 1 1 1 1 0 0 0 0 0 1
> 1 1 2 2 2 1 1 0 0 0 0 0 1 1 1
>
> IDL> bndry = find_boundary(regCat1, XSize=5, YSize=5, Perimeter=perim)
> IDL> print, perim
> 2.0000000
> IDL> print, bndry
> 0 0
> 0 1
Hello,
I think David's FIND_BOUNDARY is looking for the 1-dimensional
subscripts of the 'blob' within the image. Try...
index = where(arrayTest EQ 1, cnt)
If cnt GT 0 Then $
bndry = find_boundary(index, XSize=5, YSize=5, Perimeter=perim) else
bndry = -1
Be careful using LABEL_REGION, it assign features that lie along the
image edge to the background.
Ben
Ben
|
|
|
|
| Re: find_boundary routine [message #38669 is a reply to message #38636] |
Wed, 24 March 2004 10:29 |
brian_frizzelle
Messages: 2
Registered: March 2004
|
Junior Member |
|
|
David and Ben,
Thanks for the assistance. I now understand how David's routine works,
and it attacks this perimeter problem in a different manner than what
I needed. So I wrote a different routine which will calculate the
boundary of a region in an array, as long as all cells in the region
are coded with a 1.
The reason that the FIND_BOUNDARY routine didn't work for me is that I
needed to calculate the perimeters from the pixel corners rather than
the centers. So a one-pixel region should have a perimeter of 4, and a
2-by-2 region should have a perimeter of 8. In your routine, the
one-pixel region's perimeter would be 0, and the 2x2 pixel region's
perimeter would be 1.
If anyone is interested, I can post my code.
Brian
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
