How can I undo an intentional Gaussian blur when I don't have the originals?

Question

I have a bunch of old pictures which I had performed a Gaussian blur on to use in a competition years ago and have now realised I do not own the originals anymore. Does anyone know of a tool out there which can help me sharpen the images to be able to at least identify what the pictures were of? I'm highly curious as I do not remember where I took these pictures. All that I know is that they were pictures of landmarks in my country.

Answer 1

If you remember the exact radius of the Gaussian blur, and you processed and saved the images in a 16 bit or greater format then you can remove the blur by inverse filtering in Matlab.

If you don't remember the radius, or you truncated the pixel values by working or saving the image as 8 bit (and then used lossy JPEG compression for good measure destroying data further) then standard inverse filtering will fail spectacularly. You will get a slight improvement using a Weiner filter, or other commercial blur reduction tool, but if they're so blurred you can't tell what the pictures are of, then you're out of luck - the information you seek is no longer contained in the images.

Answer 2

Richardson-Lucy deconvolution with a Gaussian point-spread function is another algorithm which will undo your blur nicely.

You'll need to know your original radius and be able to plug it in to the PSF. Here is a list of software (free and commercial) which supports deconvolution. You'll have to try a few to discover which will let you plug in the parameters you need to supply.

Hope that helps. Good luck!