Enhancing old journal for text recognition

I wonder if photographing the pages with ultra violet or infra red might help.

I gave it a try with the attached sample. First I tried some Photoshop layer techniques, including a radial Gradient fill layer to even out the hotspot. But I wasn’t quite happy with it. I remembered that some of the most powerful features are not in Photoshop but in Camera Raw, so I started over by choosing the command Filter > Camera Raw Filter and using its features to try to bring back the type. You can see the settings I used in the pictures below:

• The main approach was to use the Light panel to optimize overall contrast, especially for the faint text.
• The Color options were adjusted to try to remove the color cast, but in the end I decided that for the goal of trying to read the handwriting, it was OK to just take out the color by slamming Saturation to -100. 
• The Effects panel options are largely about local contrast at different frequencies (except for Dehaze which is a different model), so I thought those would be useful in trying to enhance the faint text.
• I applied a radial Gradient mask to even out the hot spot that almost washes out the text.

Starting from that scan, this is really tough and I’m still not sure you can make out the faintest words.

Before/after: The image is converted to a Smart Object so that Camera Raw Filter settings can be applied nondestructively (allowing edits later).

Camera Raw Edit settings (left) that apply to the entire image, and Mask settings (right) for the radial gradient mask that tries to compensate for light falloff:

The gradient to darken the faintest text in the hot spot could also be done in Photoshop, using a radial Gradient Fill layer. That version is shown below, just so you know how that could be set up. The basic idea here is that the Overlay blending mode is applied to the radial gradient fill layer. With Overlay, 50% gray is neutral, black darkens, and white lightens, so you just add the gradient stops you need and set the gray tone of each stop to the amount of lightening or darkening you need at that distance. You can see that for this radial Gradient Fill layer, the center is set to black (darken), the edge is set to 50% gray (no change), and the gradient midpoint is shifted slightly to the left (I might have moved it a little too far).

I looked into the CZUR scanner you used, and I think I see part of the challenge here. One issue is that it uses a point light source (lamps in the suspended scanning head); like a street lamp it will have a bright center and dim towards a circular edge. That might be what causes the uneven lighting where it’s bright at left center and dims toward the right corners. On top of that, tf the CZUR scanner can save only a JPEG at 8 bits per channel, that limits how easily you can separate the faint handwriting from the non-white paper.

I’m sure a CZUR scanner is more than good enough for traditional high-contrast black print on white paper, but for challenging low-contrast cases where you want to try and read faint writing on non-white paper, it would be better to scan those items with a scanner that uses a uniform light source such as a flatbed. If it was me, I would also use advanced scanning software to carefully set the exposure to make sure neither shadows nor highlights are clipped, and that it’s saved at 16 bits per channel color. That would provide more data to push and pull in Photoshop to try and make that faint text more readable.

(A well-funded museum that had to read that text might try methods such as using equipment that can 3D-scan through the paper and tell the difference between the chemicals in the paper and the writing, or at least that’s the impression I get from watching science shows on TV 🙂 )

The next question is automation. If you have a lot of these pages, it would take forever to set this up for each one by hand. If a bunch of JPEG scans were opened into Adobe Camera Raw from Bridge, it would be possible to apply these edits as a preset to many selected images at once, and then all you would have to do is check each image and adjust the settings and mask position for any image-to-image variations. (This workflow would omit Photoshop entirely, doing all of the bulk processing and editing from Camera Raw hosted by Bridge. Doing this using Photoshop layers would involve writing an action or script that would open each document to add the right layers, which would take a lot longer and use a lot more storage space for the files.)

As @c.pfaffenbichler has already said: a (very) good scan is the basic prerequisite for optimum image processing.

Inquire in your neighbourhood whether someone has a better scanner available. If not, then with the current state of photographic technology, professional RAW images in a light tent under an anti-reflective glass plate could possibly provide a good basis

However, even the best image processing can't conjure up what your scanner has ‘swallowed’ in terms of information.

I would recommend scanning in 16bit with some back-backing – and not using the jpg-format. 

Can you provide a sample image with non-lossy-compression?