How to generate theoretical PSFs for 2D deconvolution ? (DeconvolutionLab2 PSFgenerator)

Hi,

I’m trying to generate a theoretical PSF for deconvolution of 2D images. Most of the tools I have seen (e.g. PSFGenerator plugin from @daniel.sage) seem to generate 3D PSFs.

So my current thinking is to generate a 3D PSF with a Z-depth approximating the depth of field for the given NA/magnification and then take a sum/mean Z-projection of the PSF for 2D deconvolution.

Is that a valid approach or is anyone aware of a tool that directly models PSFs for 2D deconvolution?
Tagging
@bnorthan

You can always take only 1 slice of the 3D PSF!

When every steps is well done, the deconvolution can enhance small structures and mainly decrease the out-of-focus by taking into account the axial (Z) information. This only happens in 3D - with 3D data and 3D PSF. In 2D, you can only expect to have denoising and some increase of contrast.

2 Likes

Thanks for your promt reply @daniel.sage

I yesterday tried a mean z projection and got good deconvolution results for the PSF.
I will try with a single slice as well, as per your suggestion.

Totally agree on the improvements you can get by using 3D, however I am doing this for a facility user who captured lots of images in 2D already (before asking about deconvolution) so I’ve got to work with what they have.

BTW, thanks a lot for DeconvolutionLab and the accompanying paper, it is a super useful tool for interactively experimenting with the different algorithms and PSFs (I then use flowdec for the actual batch deconvolution).

1 Like

I agree with @daniel.sage. Taking one slice of the 3D PSF should work, if the sample is in focus, or you know the z distance from “focus”.

This will work if the sample you are imaging only contains objects that are well contained within the Depth of Field of the instrument, or in other words there is not much variation in z-location of the objects.

If there are objects that vary in z-location, then there will be blur from objects out of focus that is hard to remove.

1 Like