Hi Neeraj,
I've also used the XCOSM (Expectation Maximization) algorithm in the past I
found it gave very nice results, at least on my Images.
I never managed to get it original implementation to run properly on either
linux or a mac (http://bigwww.epfl.ch/vandeville/XCOSM/). So what i used instead
was a program called ImageTrak (developed by Peter Stys)
which includes the XCOSM algorithm for deconvolution of images.
ImageTrak comes packaged as a typical mac .dmg install and is reasonably user
friendly. Unfortunately 1: the program only works on power-pc macs (e.g. G5's)
and 2: the original website is down and the program seems not to be available
any more on the internet.
If you have a PPC mac available and want to give it a try, contact me of list, I
should be able to dig out a copy of the program.
Recently I've mostly switched over to using the ImageJ Parallel Iterative
Deconvolution plugin from Piotr Wendykier mentioned by Johan, which at least for
me gives much better results than some of the earlier ImageJ Deconvolution
plugins. It also has the advantages of 1: quite fast on multi-cpu machines, I
don't bother cropping my images anymore.
2: the latest version can be called from a macro, so I can write a macro to
batch-process my images.
in my hands the FFT preconditioner causes problems, so i disable it, and I
usually use the HyBR algorithm (although they all give my pretty similar
results, as you would hope!).
I hope this helps,
manu
Quoting Stanislav Vitha <[log in to unmask]>:
> Some time ago I used XCOSM http://www.essrl.wustl.edu/~preza/xcosm/
> under linux (with graphical user interface), and also in command-line mode on
>
> a Windows machine with cygwin installed.
> I did not do any extensive tests, for my few samples I was quite satisfied
> with
> the results. The best algorithm (Expectation Maximization) has very slow
> convergence, so you need hundreds or thousands of iterations (= a lot of
> time).
> I liked the results better than those from the commercial AutoDeblur v. 9.2
> package (Blind deconvolution), but on the other hand, there the deconvolution
>
> took about 1/10th of the time.
>
> Sincerely,
> Stan Vitha
>
> Microscopy and Imaging Center,
> Texas A&M University
>
> :
> On Wed, 26 Nov 2008 00:35:13 +0100, Johan Henriksson
> <[log in to unmask]> wrote:
>
> >Neeraj Gohad wrote:
> >> Dear List,
> >>
> >> We are trying to compare open source deconvolution algorithms with the
> >> commercial deconvolution package that we have. So far we have tried the
> >> Iterative Deconvolve 3D for ImageJ. I wanted to know which other open
> >> source algorithms people are using for deconvolution.
> >>
> >> Best Regards,
> >>
> >
> >haven't tried yet:
> >http://piotr.wendykier.googlepages.com/iterativedeconvolution
> >
> >also used your package there, it failed, and since imagej has problems
> >with batching I wouldn't bother with IJ plugins for it anyway. I have
> >used my own code in the past, but for special classes of deconv, not the
> >normal one in biology. we're writing a new algorithm at the moment.
> >
> >/Johan
> >
> >--
> >--
> >------------------------------------------------
> >Johan Henriksson
> >MSc Engineering
> >PhD student, Karolinska Institutet
> >http://mahogny.areta.org http://www.endrov.net
>
--
Manu Dubin
Abt. Genetik
Univ. Kassel
Heinrich-Plett-Str. 40
34132 Kassel, Germany
phone: (++49) 561 8044178
fax: (++49) 561 8044800
[log in to unmask]
http://www.biologie.uni-kassel.de/genetics/
--------------------------------------------------------
This mail sent through http://www.uni-kassel.de/www-mail
|
