I'm aquequate at best using Gimp, so for all I know this may be something easy to do.

I'm dealing with really old church documents that were photographed in black and white, then scanned. I cannot improve the source material by taking a color photo, etc. But I've found that by playing with the filters in Gimp, I can often improve the readability.

And then there's these really problematic documents. There's 20+ pages from the end of the 1500s where it looks like they used a technique common at the time for stretching ink - adding water to it. This made the ink last longer, but the text also tended to bleed quite badly as time went on. So instead of this:

enter image description here (1624)

We instead get this:

enter image description here (1594)

The text leeched through the page a little more than when water was not added, but much more problematic, the lines themselves spread out over time, blurring into nearly unreadable messes. Here's an example of the bottom and top of two consecutive pages that had this problem:

enter image description here

Playing with the highlights in Gimp, there definitely seems to be some useable info here to play with:

enter image description here

This specific example is the marriage record of my great-great-great-great-great-great-great-great-great-great-great-grandparents. Unfortunately, while I can read the groom's name, I cannot read the name of his bride, and it's the only document on which her name would have been written. With a good deal of difficulty, the rest has been able to be read:

"Item Emerich Braunschweigs und ____ ____...
...____ zusammen gegeben den 13 decembris"

Can anyone suggest anything that might be done to help get documents like this into any more of a readable shape? In my perfect world, I imagine that I could somehow reduce the blurred lines entirely back down to just the lines made by the nib of the pen at the time, though I know that's asking far too much.

  • 2
    GIMP and other photo editing software can enhance color and contrast. But that's useless in cases where text has degraded in a complex way where each letter is covered by different pattern of ink or dirt. You need something which understands the happened degradation mechanism, the original writing style and procedure and can reason backwards the most probable original form. That's something the makers of CSI tv-series have imagined and Artificial Intelligence programmers may be able to make real. I'm afraid that the original paper must be analyzed, not only a single photo of it.
    – user82991
    Sep 30, 2021 at 6:33
  • 2
    (continued) Academic researchers who try to decipher the history may have better answers. I guess they have already met the same problem and know more than most of us, who mostly create new images. At least they have tools which shoot photos by using much finer spectral resolution and wider wavelength range than ordinary cameras and scanners.
    – user82991
    Sep 30, 2021 at 6:46
  • The problem here looks like the poor quality of the photographs to be honest, and the fact they're black and white isn't helping. There may not be much you can do with these images to be honest. Garbage in = garbage out. Doesn't matter what image editor you use, there are limits on what can be done if the image quality is bad to begin with. It would have probably been better if the images had been colour, at least then you'd have had more channels to work with.
    – Billy Kerr
    Sep 30, 2021 at 10:55
  • I agree, color and higher resolution would be helpful. But is there anything that could be done, perhaps some filter that progressively shrinks the black's outlines by a pixel or something?
    – BrianFreud
    Sep 30, 2021 at 13:11
  • Maybe try Filters > Generic > Dilate. It might help a little. You might find other possible filters in the G'MIC plugin for GIMP, some of the sharpening/repair filters in the plugin might help.
    – Billy Kerr
    Sep 30, 2021 at 15:19


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.