DeconvolutionLab in ImageJ (Fiji)

[bcli]

Hi all!

I am currently in the early stages of implementing deconvolution to images acquired through epi-fluorescence microscopy.

I have imaged subresolution beads that can be used as point spread functions that describe the diffraction patterning under its specific image acquisition conditions (wavelength etc). Theoretically, I should be able to use an averaged PSF and deconvolve a single subresolution bead and reconstruct it into a bead of higher resolution (without the diffraction blur). Image dimensions and Z steps are all the same for my images (Z steps enough to visualize clear diffraction blur).

I have mainly been using the Richardson-Lucy algorithm in DeconvolutionLab, but my images have been coming out identical to my input image. I ensured the inputted images are correct as well, and I believe I have tried close to everything with respect to fiddling with deconvolutionlab. Perhaps it is a problem intrinsic to the bead images collected?

Has anyone else encountered the same problem?



Appreciate the input and thank you in advance!

Warmly,
Brian

[bnorthan]

Hi Brian

Is your output image "exactly" the same??  Or just close??

Richardson Lucy is a slowly converging algorithm.  So you might need to run
it for a couple hundred iterations or more.  The default is 10 for some
reason.... but it is unlikely to converge in 10 iterations.

Also (if you haven't all ready) make sure you select 'flip PSF quadrants'
on the PSF options if your PSF is centered.

Finally make sure the image you want to deconvolve is the "active image"
before you run the algorithm.  I don't think there is an option to
explicitly select the input... it just uses the active image.

If you are still having problems maybe you could share some of your
images??  That would make it easier for other people to trouble shoot.

Brian

On Sun, Nov 30, 2014 at 9:43 PM, bcli <[hidden email]> wrote:

> Hi all!
>
> I am currently in the early stages of implementing deconvolution to images
> acquired through epi-fluorescence microscopy.
>
> I have imaged subresolution beads that can be used as point spread
> functions
> that describe the diffraction patterning under its specific image
> acquisition conditions (wavelength etc). Theoretically, I should be able to
> use an averaged PSF and deconvolve a single subresolution bead and
> reconstruct it into a bead of higher resolution (without the diffraction
> blur). Image dimensions and Z steps are all the same for my images (Z steps
> enough to visualize clear diffraction blur).
>
> I have mainly been using the Richardson-Lucy algorithm in DeconvolutionLab,
> but my images have been coming out identical to my input image. I ensured
> the inputted images are correct as well, and I believe I have tried close
> to
> everything with respect to fiddling with deconvolutionlab. Perhaps it is a
> problem intrinsic to the bead images collected?
>
> Has anyone else encountered the same problem?
>
>
>
> Appreciate the input and thank you in advance!
>
> Warmly,
> Brian
>
>
>
> --
> View this message in context:
> http://imagej.1557.x6.nabble.com/DeconvolutionLab-in-ImageJ-Fiji-tp5010704.html
> Sent from the ImageJ mailing list archive at Nabble.com.
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[bcli]

Hi Brian,

Thank you for the input. Unfortunately, I have already tried all of your suggestions (used 200 iterations in R-L algorithm) and the output is exactly the same! I feel like the problem lies within my collected images; the image uploaded is a centered version of the original and it is attached that is further cropped down to 100x100 pixels just for a faster upload; my original image is 512x512 pixels.

I'm a second year undergraduate student volunteering at a lab and, thus far, my project has been a bit troublesome to say the least.


Appreciate the advice.

PSF-sample_100x100pixels.tif

[bcli]

I just deconvolved using my cropped image and I was able to reconstruct a nice looking bead. I'm not too sure why that is but I'll look into it!

Warmly,
Brian