Deconvolution answers from previous discussion

Previous Topic Next Topic
classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view

Deconvolution answers from previous discussion

Hi Ash

After seeing your question a few minutes ago, I just remembered you had
sent me some other questions about deconvolution... I figure the answers
may be of interest to the group so am posting them on the list serv...

in richardson lucy algorithm, are they taking into consideration the noise
> generated from ccd camera???

Richardson Lucy is derived under the assumption of Poisson Noise.  Even
so it is well known that the noise may be amplified if you run the
algorithm for too many iteration.  Thus the algorithm is often further
modified for better noise handling.  For example both DeconvolutionLab and
ImageJ-ops deconvolution imploement Richardson Lucy with Total Variation.

is the PSF of microscope generated based on poisson or gaussian noise??

The PSF is not generated with noise.  PSF is either generated with a
diffraction based equation, or measured.  In the second case it may  be
noisy, so usually one averages several PSFs or uses PSF distilling a la

can you please tell me the steps that you followed to do deconvolution/??
> i havint completely got the logic of normaliuze psf and flip psf quadrants
> if active!!!!

In deconvolution lab, the flip psf quadrants, moves the center of the PSF
to location 0,0 as internally the algorithm needs it that way.   Normalize
PSF, makes the sum of the PSF 1.  This is needed, because otherwise you
will not preserve the energy in the image.

I also have written algorithms for deconvolution in imagej-ops, including a
method to extract the PSF from beads.  There is an active discussion about
the algorithms and related scripts here


ImageJ mailing list: