Suppose I have an image A. Is it possible to construct an image A' from A so I can get the original A after applying gaussian blur (or other convolution kernels) on A'? If it's possible, how to do it?
$\begingroup$
$\endgroup$
1
-
2$\begingroup$ in the frequency domain, you would divide by the Gaussian filter response; then applying the Gaussian blur would get you back to the original image; there are issues with the stability of the procedure, but this is the basic idea. $\endgroup$– Carlo BeenakkerCommented Sep 14, 2022 at 20:53
Add a comment
|