Newsgroups: sci.image.processing
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!bloom-beacon.mit.edu!gatech!swrinde!pipex!uunet!comp.vuw.ac.nz!canterbury.ac.nz!otago.ac.nz!bifrost.otago.ac.nz!nmein
From: nmein@bifrost.otago.ac.nz (Nick Mein)
Subject: Watershed algorithm shortcut?
Message-ID: <D05vxE.Iyo@news.otago.ac.nz>
Sender: usenet@news.otago.ac.nz (News stuff)
Nntp-Posting-Host: bifrost.otago.ac.nz
Organization: University of Otago
X-Newsreader: TIN [version 1.1 PL8]
Date: Fri, 2 Dec 1994 02:22:26 GMT
Lines: 33

The watershed algorithm (Vincent & Soille, 1991, p589) contains the
step:

  for every pixel p' which is a neighbour of p {
    if imd(p') < current_dist && (imo(p') > 0 || imo(p' == WSHED) {
      /* ie, p' belongs to an already labelled basin or to the watersheds */
      if imo(p') > 0 {
        if imo(p) == MASK || imo(p) == WSHED then
          imo(p) = imo(p');
/****/  else if imo(p) != imo(p') then
/****/    imo(p) = WSHED;
      }

On page 590:

"At this point, to get rid of the WSHED-pixels [...] it suffices to give
them the value of one of their labelled neighbors".

My question is this: If the WSHED pixels are not required, is it
possible to omit the lines /****/, and thus never create WSHED pixels in
the first place?

I've tried doing so, and get good results, but would be interested in
any comments on how valid this is.

Cheers,

Nick Mein.


Ref:
Vincent, L., & Soille., P (1991) "Watersheds in digital spaces: An efficient
   algorithm based on immersion simulations" IEEE Trans PAMI, 13, 583-598.
