Hi Andres,

Thanks for the posting.  Very informative, and I'm glad someone is doing
this.  However, I'm not sure I believe that everything that I'm seeing
reflects real life.  The images have an xy pixel size of 80nm, correct?
Most of the images seem to be "pixelated" at a lower frequency, by roughly a
factor of 3, or 240 nm.  I see this even in the original image, before
deconvolution.  I'm wondering if your calculation of the simulated bead
image, before the noise is added, limits changes in intensity across x and y
to the optical resolution (260 nm).

I would rather see real beads.

Cheers,
Jeff
Jeff Reece
Biomedical Engineer
Confocal Microscopy Center
National Insitute of Environmental Health Sciences
P.O. Box 12233, MD F2-02
Research Triangle Park, NC  27709

Ph; (919) 541-0311
Fx: (919) 541-1898
Email:  [log in to unmask]


> ----------
> From:         Andres
> Kriete[SMTP:[log in to unmask]]
> Reply To:     Confocal Microscopy List
> Sent:         2000/03/28 11:31 AM
> To:   [log in to unmask]
> Subject:      Re: Deconvolution software
>
> Dear all -
>
> seeing is believing ... as Mario Moronne pointed out. So I decided to
> put some of the images onto my webpage, which I presented at the
> FOM 1999 in Heidelberg to contribute to the ongoing discussion.
>
> We have compared 7 deconvolution methods. In order to compare,
> we used a simulated data sets of beads, convoluted with a PSF and
> added with noise - so we know what to expect (this you never know
> at 'real' specimen).
>
> Please compare our results at http://www.med.uni-giessen.de/ipl
> which you find it under the deconvolution topic. May be this is an
> initiative to find a common method to evaluate quality.
>
> Please also check for my comments added to that page. We have to
> be aware of a number of freedoms, the developers and the users have
> when operating with deconvolution software - this is the more difficult
> part.
>
> Andres
>
> ---------------------
> Priv.-Doz.Dr.Andres Kriete
> Image Processing Lab
> University Clinic Giessen
> Aulweg 123
> 35385 Giessen - Germany
>
>
>
>
> Joachim Walter wrote:
>
> > Probably the whole topic of what deconvolution algorithm / software
> > package works best in which situation could be cleared up a lot if the
> > people working in that field could agree on a common method of
> > evaluating the quality of the results. I heard that there have been a
> > few attempts to measure the quality of image restoration. But it seems
> > none is generally accepted in the sense that e.g. software vendors would
> > quote the results to advertise their software.
> > Is there any change in this to be expected?
> >
>