How can I clean up JPEG compression artifacts in a scanned book image using GIMP?

Asked 5/25/2016

1 views

2 answers

0

I photographed or scanned a page from a book, and the image has lots of speckled gray artifacts around the text and lines. This does not look like typical sensor noise. I’d like to clean it up in GIMP if possible, ideally without manually retouching the whole page. What tools or adjustments work best for this kind of artifacting?

Originally by Photography Stack Exchange contributor. Source · Licensed CC BY-SA 4.0

Photography Stack Exchange contributor

10y ago

2 Answers

5

What you are seeing in your example is not random noise. It is what is referred to as compression artifacts and is what happens when an image is compressed too much to make the file size smaller.

Instead of trying to use noise removal, simply use a levels or curves layer. Or even try just boosting the contrast. You have a white background and dark text and lines, and the noise is a light or middle grey, so any of those methods can be adjusted to remove the lighter tones and make them white.

For example, quickly using pixlr and boosting contrast, I get this result. I'm sure you can do better with GIMP and levels or curves

enter image description here

Using a curve:

enter image description here

Originally by user4191. Source · Licensed CC BY-SA 4.0

user4191

10y ago

0

AI Answer

Generated from our catalog & community — verify before relying on it.

This looks like JPEG compression artifacts rather than random camera noise. Because your page is mostly dark text on a white background, a noise-reduction filter is usually not the best fix.

In GIMP, try tonal adjustments instead:

  • Levels: move the white point inward so the light gray speckles become pure white.
  • Curves: lift the highlights and increase contrast to separate the background from the text.
  • Contrast: a simple contrast boost may already reduce much of the artifacting.

Why this works: the unwanted artifacts are mid/light gray, while the page background should be white and the text is much darker. By pushing light tones to white, you can remove much of the visible mess without harming the text.

If needed, combine this with a careful threshold-style adjustment, but start with Levels or Curves first—they give more control and are less likely to damage fine detail.

The best long-term solution is to start from a less-compressed source file if possible, since heavy JPEG compression permanently throws away image information.

UniqueBot

AI

10y ago

Your Answer