| Re: Efficient finding of locations where two 'on' pixels are next to one another [message #74359] |
Thu, 13 January 2011 10:21 |
Marc Buie
Messages: 12
Registered: December 2010
|
Junior Member |
|
|
Robin -
I wrote a very similar routine some time ago. The name is collapse.pro and is in my library (www.boulder.swri.edu/~buie/idl). I worked really hard on this to speed it up and it's pretty good though you might not believe this to see the code. My program might be different from yours though. I use two images, one a bit image that is being processed but it is coupled to another "grey-scale" image from which the bit image is derived. I work with astronomical data and my images are of the sky (with stars and such) and the bit image records the positions of the stars.
I use boxm.pro which finds a local max in the image and uses that information to turn off the bits in the bit image that are too close to one another.
You might find some useful tricks in these two programs that could help you solve your problem.
Cheers,
Marc
|
|
|
|
| Re: Efficient finding of locations where two 'on' pixels are next to one another [message #74360 is a reply to message #74359] |
Thu, 13 January 2011 09:56  |
jeanh
Messages: 79
Registered: November 2009
|
Member |
|
|
On 13/01/2011 12:28 PM, Robin Wilson wrote:
> Hi,
>
> I'm working with a binary image which is the output of another procedure
> I've run. I am trying to write a routine to post-process this image, and
> clean it up a bit. As part of this, I want to find the indices of pixels
> that are set to 1 (ie. 'on') that are next to other pixels that are set
> to 1, but only in the X direction. To make that a bit clearer, there is
> a diagram below:
>
>
> (0) . . . * * . .
> (1) . . . * . . .
> (2) . . * * * . .
> (3) . . . * . . .
> (4) . . . * . . .
>
> In this example I would want the indices of both 'on' pixels in Row 0,
> nothing in Row 1 (as the pixel is not touching anything in the X
> direction), all three in Row 2, and nothing in Rows 3 and 4.
>
> At the moment all I can think of some kind of ghastly FOR loop with
> loads of state-checking. I'm thinking that there must be a nicer (that
> is more efficient and cleaner code) way...is there?
>
> Cheers again,
>
> Robin
You can try using "shift"
pix = where(img eq 1 and shift(img,-1,0) eq 1)
pix = [pix,pix+1] ;include the index of all valid pixels
pix = pix[uniq(pix,sort(pix))] ; remove duplication caused by the
previous line
You handle the behavior at the edge
Jean
PS: untested, double check that I didn't got confused between +1 and -1
in the shift()
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
