I'd be happy to include your suggested changes, if it's simpler to go that route than have two separate pull requests.  I don't want to step on any toes regarding attribution, though.  

Having the 'soft' probabilities available is definitely a useful improvement (cleanup, fuzzy post-processing), and I'm glad you implemented that!  


On Monday, August 20, 2012 4:11:45 PM UTC-5, Emmanuelle Gouillart wrote:
Hi Josh,

extending the random walker algorithm to multichannel images sounds like
a very nice idea, and I'm sure that it would be useful for other people
(including myself!).

In fact, I had started working on some improvements of the random walker
code but didn't go as far as proposing the pull request. I have a branch
for that on
https://github.com/emmanuelle/scikits.image/tree/fix_random_walker, maybe
you can include these changes when you contribute your changes?


On Mon, Aug 20, 2012 at 10:03:53AM -0700, Josh Warner wrote:
>    I have modified the existing random walker algorithm into a fully
>    backwards-compatible version which allows inclusion of multispectral data,
>    e.g. RGBA channels or different (registered) image modalities. �I really
>    liked the existing algorithm, so I just extended it rather than write one
>    from scratch for my own purposes. �The overhead is minimal; multispectral
>    processing is triggered if data is passed as an iterable of arrays rather
>    than just an array. �
>    This amounts to combining image gradients as sqrt(sum-of-squares) and
>    dividing by sqrt(#channels). �For obvious reasons, the several channels
>    must be pre-processed to have data on similar ranges by whitening or a
>    similar method. �Not usually a problem for RGB, but in medical imaging
>    this rears its head.
>    Would this be of interest to the community? �I'd be happy to contribute
>    the changes if there is interest.