Parallel Spectral Deconvolution plugin update

[Piotr Wendykier]

Hello,

I have updated the Parallel Spectral Deconvolution plugin.
The bug causing poor results in all 3D methods with reflexive
boundary conditions has been fixed. Moreover, the source code
has been significantly simplified and the plugin now uses Parallel Colt 0.2.

The plugin is available at:

http://piotr.wendykier.googlepages.com/deconvolution

Regards,

Piotr Wendykier

[lechristophe]

I went to the plugin homepage, it seems very interesting but it didn't
explained what is the use of such a decouvolution... Could someone explain
what exactly is this plugin doing ? I'm interested in 3D deconvolution of
widefield fluorescence Z-stacks, could this be used for this application or
is it a totally different kind of deconvolution ?

On Jan 7, 2008 12:10 AM, Piotr Wendykier <[hidden email]> wrote:

> Hello,
>
> I have updated the Parallel Spectral Deconvolution plugin.
> The bug causing poor results in all 3D methods with reflexive
> boundary conditions has been fixed. Moreover, the source code
> has been significantly simplified and the plugin now uses Parallel Colt
> 0.2.
>
> The plugin is available at:
>
> http://piotr.wendykier.googlepages.com/deconvolution
>
> Regards,
>
> Piotr Wendykier
>

[Piotr Wendykier]

Dear Christophe,

The short answer is:  yes, it could be used for your Z-stacks,
and no it is not a totally different kind of deconvolution.

The long answer is harder to explain.  There is just one definition of
deconvolution, but there are a lot of different methods that can be used
to actually do the deconvolution. Furthermore, some people use different
names for essentially the same methods. Why different methods? That's
hard to explain without using a lot of mathematics. The basic idea is that
the result produced by deconvolution algorithms depends on the image, the
noise in the image, and the PSF. One deconvolution method might work well
for one problem, but not so well for another problem. So it is good to
try different methods. This plugin uses fairly standard deconvolution
methods, with two advantages. First, we provide different deconvolution
methods that you can try with your data -- maybe one will work better
than another. Second, our methods have been implemented to be very
efficient on modern multi-core processor machines.
If you want to know more about these algorithms, I recommend an excellent
book: "Deblurring Images: Matrices, Spectra, and Filtering" by
Per Christian Hansen, James G. Nagy, and Dianne P. O'Leary, SIAM 2006.

Piotr

On Jan 7, 2008 4:36 AM, Christophe Leterrier
<[hidden email]> wrote:

> I went to the plugin homepage, it seems very interesting but it didn't
> explained what is the use of such a decouvolution... Could someone explain
> what exactly is this plugin doing ? I'm interested in 3D deconvolution of
> widefield fluorescence Z-stacks, could this be used for this application or
> is it a totally different kind of deconvolution ?
>
>
> On Jan 7, 2008 12:10 AM, Piotr Wendykier <[hidden email]> wrote:
>
> > Hello,
> >
> > I have updated the Parallel Spectral Deconvolution plugin.
> > The bug causing poor results in all 3D methods with reflexive
> > boundary conditions has been fixed. Moreover, the source code
> > has been significantly simplified and the plugin now uses Parallel Colt
> > 0.2.
> >
> > The plugin is available at:
> >
> > http://piotr.wendykier.googlepages.com/deconvolution
> >
> > Regards,
> >
> > Piotr Wendykier
> >
>

[Justin McGrath]

How do you determine the PSF?  I was interested in deconvolving a
fluorescence image but wasn't sure how to get the PSF of the
microscope.  Just for grins I cut and pasted an example of an Airy
disk from a journal article and used that.  To my surprise, it worked
pretty well, but there has to be a better way.

Since I know about the optics of the microscope can I calculate the
PSF?  If so, can you point me to some place that explains how?

Justin

On Jan 7, 2008 7:26 AM, Piotr Wendykier <[hidden email]> wrote:

> Dear Christophe,
>
> The short answer is:  yes, it could be used for your Z-stacks,
> and no it is not a totally different kind of deconvolution.
>
> The long answer is harder to explain.  There is just one definition of
> deconvolution, but there are a lot of different methods that can be used
> to actually do the deconvolution. Furthermore, some people use different
> names for essentially the same methods. Why different methods? That's
> hard to explain without using a lot of mathematics. The basic idea is that
> the result produced by deconvolution algorithms depends on the image, the
> noise in the image, and the PSF. One deconvolution method might work well
> for one problem, but not so well for another problem. So it is good to
> try different methods. This plugin uses fairly standard deconvolution
> methods, with two advantages. First, we provide different deconvolution
> methods that you can try with your data -- maybe one will work better
> than another. Second, our methods have been implemented to be very
> efficient on modern multi-core processor machines.
> If you want to know more about these algorithms, I recommend an excellent
> book: "Deblurring Images: Matrices, Spectra, and Filtering" by
> Per Christian Hansen, James G. Nagy, and Dianne P. O'Leary, SIAM 2006.
>
> Piotr
>
> On Jan 7, 2008 4:36 AM, Christophe Leterrier
>
> <[hidden email]> wrote:
> > I went to the plugin homepage, it seems very interesting but it didn't
> > explained what is the use of such a decouvolution... Could someone explain
> > what exactly is this plugin doing ? I'm interested in 3D deconvolution of
> > widefield fluorescence Z-stacks, could this be used for this application or
> > is it a totally different kind of deconvolution ?
> >
> >
> > On Jan 7, 2008 12:10 AM, Piotr Wendykier <[hidden email]> wrote:
> >
> > > Hello,
> > >
> > > I have updated the Parallel Spectral Deconvolution plugin.
> > > The bug causing poor results in all 3D methods with reflexive
> > > boundary conditions has been fixed. Moreover, the source code
> > > has been significantly simplified and the plugin now uses Parallel Colt
> > > 0.2.
> > >
> > > The plugin is available at:
> > >
> > > http://piotr.wendykier.googlepages.com/deconvolution
> > >
> > > Regards,
> > >
> > > Piotr Wendykier
> > >
> >
>

[Philip Ershler]

Hi,
        We determined the actual PSF for our system. We did this by  
suspending fluorescent beads in agar. Then we scanned the beads. The  
resulting Z-stacks were used to calculate the PSF for our system with  
a given objective. I was not involved in the actual process of  
manipulating the image data to produce the final PSF, but I could put  
you in touch with the responsible individual if your are really  
interested in pursuing this method.

Phil

On Jan 7, 2008, at 12:48 PM, Justin McGrath wrote:

> How do you determine the PSF?  I was interested in deconvolving a
> fluorescence image but wasn't sure how to get the PSF of the
> microscope.  Just for grins I cut and pasted an example of an Airy
> disk from a journal article and used that.  To my surprise, it worked
> pretty well, but there has to be a better way.
>
> Since I know about the optics of the microscope can I calculate the
> PSF?  If so, can you point me to some place that explains how?
>
> Justin
>
> On Jan 7, 2008 7:26 AM, Piotr Wendykier <[hidden email]>  
> wrote:
>> Dear Christophe,
>>
>> The short answer is:  yes, it could be used for your Z-stacks,
>> and no it is not a totally different kind of deconvolution.
>>
>> The long answer is harder to explain.  There is just one definition  
>> of
>> deconvolution, but there are a lot of different methods that can be  
>> used
>> to actually do the deconvolution. Furthermore, some people use  
>> different
>> names for essentially the same methods. Why different methods? That's
>> hard to explain without using a lot of mathematics. The basic idea  
>> is that
>> the result produced by deconvolution algorithms depends on the  
>> image, the
>> noise in the image, and the PSF. One deconvolution method might  
>> work well
>> for one problem, but not so well for another problem. So it is good  
>> to
>> try different methods. This plugin uses fairly standard deconvolution
>> methods, with two advantages. First, we provide different  
>> deconvolution
>> methods that you can try with your data -- maybe one will work better
>> than another. Second, our methods have been implemented to be very
>> efficient on modern multi-core processor machines.
>> If you want to know more about these algorithms, I recommend an  
>> excellent
>> book: "Deblurring Images: Matrices, Spectra, and Filtering" by
>> Per Christian Hansen, James G. Nagy, and Dianne P. O'Leary, SIAM  
>> 2006.
>>
>> Piotr
>>
>> On Jan 7, 2008 4:36 AM, Christophe Leterrier
>>
>> <[hidden email]> wrote:
>>> I went to the plugin homepage, it seems very interesting but it  
>>> didn't
>>> explained what is the use of such a decouvolution... Could someone  
>>> explain
>>> what exactly is this plugin doing ? I'm interested in 3D  
>>> deconvolution of
>>> widefield fluorescence Z-stacks, could this be used for this  
>>> application or
>>> is it a totally different kind of deconvolution ?
>>>
>>>
>>> On Jan 7, 2008 12:10 AM, Piotr Wendykier  
>>> <[hidden email]> wrote:
>>>
>>>> Hello,
>>>>
>>>> I have updated the Parallel Spectral Deconvolution plugin.
>>>> The bug causing poor results in all 3D methods with reflexive
>>>> boundary conditions has been fixed. Moreover, the source code
>>>> has been significantly simplified and the plugin now uses  
>>>> Parallel Colt
>>>> 0.2.
>>>>
>>>> The plugin is available at:
>>>>
>>>> http://piotr.wendykier.googlepages.com/deconvolution
>>>>
>>>> Regards,
>>>>
>>>> Piotr Wendykier
>>>>
>>>
>>

[lechristophe]

Hi Justin,

Some ImageJ plugins that could be usefull:

PSF Tools by the ETH Computationnal Biophysics Lab allows you to measure an
experimental PSF and compare it to a theoretical one:
http://www.cbl.ethz.ch/Downloads/psftool

Diffraction PSF 3D by Bob Dougherty can be used to generate a theoretical
PSF:
http://www.optinav.com/Diffraction-PSF-3D.htm

Deconvolution3D by Pierre Besson contains also a module to generate a
theoretical PSF:
http://bigwww.epfl.ch/demo/deconvolution3D/

For these tools to generate a theoretical PSF, you need to know some
parameters of your microscope setup and sample like NA, RI of mounting
medium, wavelength, etc. The usefullness of using a theoretical vs an
experimental (measured with beads) PSF has been discussed in great details
on the confocal mailing list:
http://listserv.acsu.buffalo.edu/cgi-bin/wa?S1=confocal&D=0&H=0&O=T&T=0

Briefly, having a theoretical PSF has the advantage of being noise-free,
whereas an experimental PSF is closer to your real parameters. My advice
would be to measure an experimental PSF and to derive parameters from it,
and then generate a theoretical one from these paramters, so get the best of
both worlds.

Some ImageJ plugins that could be usefull:

PSF Tools by the ETH Computationnal Biophysics Lab allows you to measure an
experimental PSF and compare it to a theoretical one:
http://www.cbl.ethz.ch/Downloads/psftool

Diffraction PSF 3D by Bob Dougherty can be used to generate a theoretical
PSF:
http://www.optinav.com/Diffraction-PSF-3D.htm

Deconvolution3D by Pierre Besson contains also a module to generate a
theoretical PSF:
http://bigwww.epfl.ch/demo/deconvolution3D/

For these tools to generate a theoretical PSF, you need to know some
parameters of your microscope setup and sample like NA, RI of mounting
medium, wavelength, etc. The usefullness of using a theoretical vs an
experimental (measured with beads) PSF has been discussed in great details
on the confocal mailing list:
http://listserv.acsu.buffalo.edu/cgi-bin/wa?S1=confocal&D=0&H=0&O=T&T=0

Briefly, having a theoretical PSF has the advantage of being noise-free,
whereas an experimental PSF is closer to your real parameters. My advice
would be to measure an experimental PSF and to derive parameters from it,
and then generate a theoretical one from these paramters, so get the best of
both worlds.


Christophe Leterrier

Postdoc
INSERM UMR641 Neurobiology of ionic channels
IFR Jean Roche - Mediterranee University
Marseille, France

> On Jan 7, 2008, at 12:48 PM, Justin McGrath wrote:
>
> > How do you determine the PSF?  I was interested in deconvolving a
> > fluorescence image but wasn't sure how to get the PSF of the
> > microscope.  Just for grins I cut and pasted an example of an Airy
> > disk from a journal article and used that.  To my surprise, it worked
> > pretty well, but there has to be a better way.
> >
> > Since I know about the optics of the microscope can I calculate the
> > PSF?  If so, can you point me to some place that explains how?
> >
> > Justin
>
>

[agus wijoyo]

Dear Forum,

I have the problem about picture size, I used Ki-67
plugin to count Ki-67 Labeling Index,however after I
load the picture (3136x2352 size) the Image J was
error (not responding). If I resize the picture will
it affect the result?
Thank you for any suggestion

Agus



      ________________________________________________________
Bergabunglah dengan orang-orang yang berwawasan, di di bidang Anda! Kunjungi Yahoo! Answers saat ini juga di http://id.answers.yahoo.com/

[Wayne Rasband]

You should not have any problems opening and working with a 3136x2352
RGB (28MB) image on a machine with at least 1 GB of RAM and 640MB
assigned to ImageJ. It is best to use Java 1.6 because ImageJ running
on Java 1.6 is able to open 8-bit and RGB images using half as much
memory (in this case, 28MB vs 56MB).

-wayne


On Jan 9, 2008, at 9:05 AM, agus wijoyo wrote:

> Dear Forum,
>
> I have the problem about picture size, I used Ki-67
> plugin to count Ki-67 Labeling Index,however after I
> load the picture (3136x2352 size) the Image J was
> error (not responding). If I resize the picture will
> it affect the result?
> Thank you for any suggestion
>
> Agus
>
>
>
>       ________________________________________________________
> Bergabunglah dengan orang-orang yang berwawasan, di di bidang Anda!
> Kunjungi Yahoo! Answers saat ini juga di http://id.answers.yahoo.com/
>

[agus wijoyo]

Thank you for your suggestion Mr Wayne, I only use 512
MB of RAM and using Java 1.6,is there any alternative
beside upgrade my computer to 1 GB of RAM, is it oke
if I resize my image? thank you



      ________________________________________________________
Bergabunglah dengan orang-orang yang berwawasan, di di bidang Anda! Kunjungi Yahoo! Answers saat ini juga di http://id.answers.yahoo.com/