4

For an assignment, The image corrupted by atmospheric turbulence. I want to deblur an image using inverse image filtering. I have done some research and it seems I need the original image for this procedure but I only have the blurred image. How can I construct the degrading function that was used to blur this image? I am not allowed to use the original image. Thank you in advance.

This is the image:

This is the original image

Ismael Padilla
  • 5,246
  • 4
  • 23
  • 35
Mert KARADENIZ
  • 61
  • 5
  • Welcome to Stack Overflow! Please take the [tour](https://stackoverflow.com/tour) and read [How to Ask](https://stackoverflow.com/help/how-to-ask). Please [edit](https://stackoverflow.com/posts/65857930/edit) the post to include your own effort into solving this problem. The latter preferably in code, this is called a [minimal reproducible example](https://stackoverflow.com/help/minimal-reproducible-example). – Bilal Jan 23 '21 at 12:06
  • 1
    See this [library](https://github.com/PyLops/pylops_pydata2020) you might find it useful. – Bilal Jan 23 '21 at 12:09
  • It seems useful but I am not allowed to use a third-party library. – Mert KARADENIZ Jan 23 '21 at 12:35
  • you are supposed to use DeepLearning, or just simple image processing? – Bilal Jan 23 '21 at 12:38
  • Just simple image processing. – Mert KARADENIZ Jan 23 '21 at 12:41
  • 2
    one suggestion in my mind is to use edge detection, and combine the edges with the blurred image to partially de-blur it, it's hard task for simple image processing under these conditions. – Bilal Jan 23 '21 at 12:47
  • The assignment question is this "Apply inverse filtering in order to restore the image ‘question_3.tif’ corrupted by atmospheric turbulence." So I cannot use other tecniques. – Mert KARADENIZ Jan 23 '21 at 12:53
  • 1
    Do you have the filtering function (the kernel) that was used to blur this picture? Or do you have some idea of what the function for simulating "atmospheric turbulence" would be? I don't see how this would be a reasonable assignment without that. – Ian Chu Jan 23 '21 at 17:45
  • Oh, there's someone else asking the exact same question. Apparently the Wiener Filter can estimate the unblur from a single picture. – Ian Chu Jan 23 '21 at 17:50
  • 1
    "I am not allowed to use a third-party library." Can you clarify this statement? Are you saying that you have to write code from scratch? Deblurring is some advanced stuff. – Jim O. Feb 20 '21 at 08:44
  • OP repeatedly points out that using the original image is not allowed. Obviously. But I feel that more information than shown here is available on the blurring filter, model and parameters. Inverting that would still be a challenge, but much less than to try to determine that info from the blurred picture. I recommend to ask teacher/tutor about available info on the blur. – Yunnosch Dec 02 '21 at 13:35
  • Does this answer your question? [Deblurring an image](https://stackoverflow.com/questions/53743561/deblurring-an-image) – Cris Luengo Dec 02 '21 at 15:26
  • short answer is "you can’t, lost information cannot be recovered". Long answer is "you may make it *look* less blurry by doing things like reinforcing edges (using edge filters like eg sobel) or other more advanced techniques (some machine learning based algorithms are very efficient in that) – tbrugere Dec 20 '21 at 10:44
  • oh and if it was not really blurred by "atmostpheric turbulence", but just by a simple gaussian filter in frequency domain, you may just apply the inverse of that transform in fourier domain – tbrugere Dec 20 '21 at 10:51

0 Answers0