I have an experimental PSF (point spread function) of an optical system for remote sensing images. However, it is asymmetrical and has some sidelobes as below.

Enter image description here

And I want to perform deconvolution to sharpen the images.

Is it possible to use this PSF as it is to sharpen the images? Or should there be some modifications and enhancements for the PSF before using it? Also, is there an available reference paper tackling such an issue?

  • $\begingroup$ Yes, it is possible to use your PSF 'as it is' for the purpose of sharpening the images; the image restoration technique required for the purpose is the Richardson-Lucy algorithm; the Wiki article on PSF indirectly refers to the Richardson-Lucy algorithm via the Deconvolution article reference. // Yes, there are available reference[s] tackling such an issue; web search on rl deconvolution gives plenty of those; you can find a variety of implementation on github and elsewhere. $\endgroup$
    – V.V.T
    Jul 30, 2023 at 7:13
  • $\begingroup$ Thank you for your reply sir, do you think RL algorithm is suitable for such case?from what I understand is that the noise in remote sensing images might not be Poisson but could be more Gaussian in nature, and RL assumes that the noise in images follows Poisson statistics. therefore, RL might not be the suitable algorithm. $\endgroup$
    – gin
    Oct 22, 2023 at 12:51
  • $\begingroup$ Yes, the original LR leverages poisson-ness of PSF statistics, and for some time it's been a popular subject to try and extend LR for use with gaussian blur. Some succeeded. On the other hand, some modifications and enhancements for the PSF before using it were developed, too. Try web search with keywords 'remote sensing images, richardson lucy algorithm'. You can hardly expect quality answers with unspecific questions like those of your post. Or is your question just recon? $\endgroup$
    – V.V.T
    Oct 23, 2023 at 13:27
  • $\begingroup$ On the other hand, it seems quite natural to assume that ill-posed problems generate ill-posed forum questions ;).Your comments lay bare your involvement in the astronomical telescopes and remote sensing. Maybe these references are not a revelation for you, but, for random visitors from web searches, A convergent blind deconvolution method for post-adaptive-optics astronomical imaging (arxiv.org/pdf/1305.0421.pdf) and An all-photonic focal-plane wavefront sensor (nature.com/articles/s41467-020-19117-w) may help explain what you are after. $\endgroup$
    – V.V.T
    Oct 25, 2023 at 5:58

1 Answer 1


PSF can be used for deconvolution, however:

  • Make double triple sure that this is the actual PSF and not a random glimpse (atmospheric speckle/seeing comes to mind)

  • If it is a momentary PSF, then it can be used only for the image taken at the same time. e.g. if it is taken from a reference section in a larger frame, the the entire frame can be deconvolved, but no other frames.

  • Consider widening it. Good results appear when you know the PSF over a large area and with a wide dynamic range and low noise.

  • $\begingroup$ thank you for your reply, it was taken from a star , as a point source. could you please explain what do you mean by " momentary PSF"? also, by widening it, you mean scaling the width of it? and do you have reference to follow your suggested approach. $\endgroup$
    – gin
    Jul 29, 2023 at 13:54
  • $\begingroup$ If it is from a star seen through atmosphere, then this PSF is not constant, because the atmosphere is not constant. It can be only applied to the same image it was taken from and the portions that are immediately next to this star. For portions further away, even in the same image, again the PSF will be different there, because you are looking through another part of the atmosphere. Widening means taking a larger piece of your image $\endgroup$
    – tobalt
    Jul 30, 2023 at 5:14
  • $\begingroup$ Thank you for the explanation sir, so in this case, do you recommend to use a blind deconvolution considering the PSF is not constant? $\endgroup$
    – gin
    Oct 22, 2023 at 12:54

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.