I'm trying to get the best possible scan of a CD cover (although this would apply to any print media using CMYK halftoning).

With the highest setting on my scanner (3200dpi) I quickly run into the DPI of the print itself.

Here's a small portion for an example:

Lots of CMYK dots making up a small piece of a guitar

I can certainly blur the image, but there doesn't seem to be a good compromise for the blur radius between still having some grid artifacts and loosing too much detail. This is my best attempt:

A blured picture of the same guitar, with visible grid patterns

Is there a better way to do this? Surely there's some filter that has more "knowledge" about halftones and can make a better result because of it.

The bluring above was done in GIMP, and while I'd love a GIMP solution, I'd appreciate any tool that can get the job done.

  • 1
    \$\begingroup\$ You can try using Fourier transforms... What is the best way to remove texture from a scanned textured photo paper? \$\endgroup\$
    – xiota
    Jan 7, 2021 at 1:22
  • \$\begingroup\$ I don't see how to use fourier transforms here. That question is about an image that has a repeating pattern "on top of" it which needs to be removed/filtered out, where in this case the pattern is the image; removing it would remove the image. This is also why something like despeckle doesn't work. If you know a way to apply fourier transforms to this problem that would make a great answer. \$\endgroup\$
    – Shelvacu
    Jan 7, 2021 at 6:46
  • \$\begingroup\$ Removing the halftone pattern will reduce the apparent resolution and sharpness of the image. There's nothing you can do about that because the detail simply isn't there. Otherwise, the procedure is exactly the same as in the other question \$\endgroup\$
    – xiota
    Jan 7, 2021 at 7:41
  • \$\begingroup\$ Same as@xiota. There is a FFT plugin for Gimp. Since it is a binary availability depends on yrou OS and Gimp version. \$\endgroup\$
    – xenoid
    Jan 7, 2021 at 11:13
  • \$\begingroup\$ The halftoning process itself sets an upper limit on the amount of detail you'll be able to retain. It's unfortunate for sure. \$\endgroup\$ Jan 10, 2021 at 5:04

1 Answer 1


Please note - it has been a while since I've worked with this kind of scanning.

I think the repeating pattern (especially visible in the wood section of the image after blurring) can be caused by interference between the the scanner's grid of pixels and the grids used in the CMYK halftone screens. The traditional way to reduce the effect would be to try rotating the CD cover on the flatbed (maybe try 15, 33 and 45 degrees) and seeing if that helps the end result (after gaussian blur). You might also want to downsize the image after the blur to try and restore sharpness and reduce the visibility of any halftone dots.

Also - looking at the R, G, B channels separately the amount of blurring seems different with each one. Using your sample pre-blur image I was able to get pretty good results by using different blur settings for each channel.

  • \$\begingroup\$ I don't think the grid artifacts are caused by interference, they're caused by the blur not being strong enough to eliminate the halftone dithering in the source. The source image clearly shows the actual dot pattern. I find that resizing my CD cover scans to about 600 pixels wide does a good job, but I have my own custom resizing software based on a Lanczos-5 filter. \$\endgroup\$ Jan 10, 2021 at 4:55

Your Answer

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

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