MATLAB Central Newsreader - Image Region -> perimeter sample

Image Region -> perimeter sample

Ahmad — Sat, 10 May 2008 20:42:02 -0400

Hi there,

Given a binary image of a single connected object, what is
the quickest way to find a sample of pixels on its
edge/perimeter. By a sample of pixels I do not mean all
pixels on the perimeter, but just a sample of fixed number
of pixels, where each pixel is separated from the other by
some margin. (I also have the Image processing toolbox)

Regards,

Re: Image Region -> perimeter sample

roberson@ibd.nrc-cnrc.gc.ca (Walter Roberson) — Sat, 10 May 2008 22:41:35 -0400

In article <g051aq$m6a$1@fred.mathworks.com>,
Ahmad <ahmad.humyn@gmail.com> wrote:

>Given a binary image of a single connected object, what is
>the quickest way to find a sample of pixels on its
>edge/perimeter. By a sample of pixels I do not mean all
>pixels on the perimeter, but just a sample of fixed number
>of pixels, where each pixel is separated from the other by
>some margin. (I also have the Image processing toolbox)

Could you expand on the bit about "where each pixel is separated
from the other by some margin" ? Margin in what sense?
If you are talking about geometric distances, then do you want
the -furthest- point that is at most that distance away, or
do you waht the -closest- point that is at least that distance away?
If you are talking about listing the perimeter points in a list
and taking every N'th entry from the list, then you need to
determine your traversal algorithm; for example if you are
coming down from the left-top of an "X" and you reach the centre,
then which of the three other branches is the next one to follow?

--
  "Do diddle di do,
   Poor Jim Jay
   Got stuck fast
   In Yesterday." -- Walter De La Mare

Re: Image Region -> perimeter sample

ImageAnalyst — Sun, 11 May 2008 16:45:35 -0400

On May 10, 4:42=A0pm, "Ahmad " <ahmad.hu...@gmail.com> wrote:
> Hi there,
>
> Given a binary image of a single connected object, what is
> the quickest way to find a sample of pixels on its
> edge/perimeter. By a sample of pixels I do not mean all
> pixels on the perimeter, but just a sample of fixed number
> of pixels, where each pixel is separated from the other by
> some margin. (I also have the Image processing toolbox)
>
> Regards,
----------------------------------------------------------------------------=
--------
Ahmad:
What is it about bwboundary that you don't know how to deal with?
This gives you the boundaries of objects. You could then take a
"sample of fixed number of pixels" from that just by taking the first
N pixels, or using rand to generate indexes to extract or lots of
other methods. What are you really after when you say your want just
a sample of the boundaries rather than all of them. What good would
that be?
Regards,
ImageAnalyst

Re: Image Region -> perimeter sample

Ahmad — Mon, 12 May 2008 13:42:03 -0400

Thanks :) Nothing is wrong with bwboundary or
bwtraceboundary...I just didn't know about them. Thanks for
pointing me in the right direction. These functions were
exactly what I was looking for :)

Thank you for all the help.

Regards,

ImageAnalyst <imageanalyst@mailinator.com> wrote in message
<f41d7a04-fb33-46b9-978d-746a147f91f6@b64g2000hsa.googlegroups.com>...
> On May 10, 4:42=A0pm, "Ahmad " <ahmad.hu...@gmail.com> wrote:
> > Hi there,
> >
> > Given a binary image of a single connected object, what is
> > the quickest way to find a sample of pixels on its
> > edge/perimeter. By a sample of pixels I do not mean all
> > pixels on the perimeter, but just a sample of fixed number
> > of pixels, where each pixel is separated from the other by
> > some margin. (I also have the Image processing toolbox)
> >
> > Regards,
>
----------------------------------------------------------------------------=
> --------
> Ahmad:
> What is it about bwboundary that you don't know how to
deal with?
> This gives you the boundaries of objects. You could then
take a
> "sample of fixed number of pixels" from that just by
taking the first
> N pixels, or using rand to generate indexes to extract or
lots of
> other methods. What are you really after when you say
your want just
> a sample of the boundaries rather than all of them. What
good would
> that be?
> Regards,
> ImageAnalyst