To get back on track, I can give an example of a 2015 e-book with a mix of really good and really bad images. It is currently $1.30 at amazon US and a free sample almost certainly has the tables that are the worst and likely has a couple of maps.
Quote:
Originally Posted by sufue
Marcus Agrippa: Right-hand Man of Caesar Augustus is a biography of...you can probably guess  ...by Lindsay Powell, published by UK specialty publisher Pen & Sword.
|
The multi-page "Chronology" and "List of Consuls" tables are scanned images with poor contrast and bad JPEG artifacts. Interestingly, the bottom of the "List of Consuls" table only takes up a few lines at the top of a page and was scanned with significantly higher resolution and better contrast, although the JPEG artifacts are little better.
Regardless of the target display, I think text should be scanned in monochrome at high resolution, say 1200dpi or more depending on scanner quality. (If the scanner is not that good, high resolution can get quite noisy.) Then the contrast should be aggressively stretched, then the image shrunk with a method that combines high resolution back and white neighboring pixels to make gray pixels. (Most web browsers do this as do many image processing utilities.)