
Carlo Traverso wrote:
I have one, Plustek OpticBook 3600, and I am very much satisfied of it, but scanning books in book mode trims away at least 1cm. in the middle, so can be used only if the margins are generous. To use it you have to open the book at 90 degrees, usually possible.
I use the same model, and am very happy with its speed; for 300 dpi images of 8vo sized books, I have clocked myself at 300 pages per hour on a book with a good binding. I don't know what software you use it with, but if you have Abbyy, you might do what I do and run it through Abbyy's interface rather than its own "book mode" interface. The Abbyy driver should capture the entire platen rather than throwing away the outer cm. My experience is that having the book only 90 degrees open eliminates much of the gutter shadow on its own, and the additional processing that "book mode" does is largely unnecessary.
However, in my experience, the limit is not scanning quality: it is print quality. OCR quality is pretty good on modern editions, but old books, often stained, and even more often with defective print, give raise to a lot of errors. Often you don't have the choice of a better print.
This can't be helped. However, the other issue that gives problematic raw OCR is that even when character recognition is good, layout detection can be poor, and sidenotes, multi-column text, and the like can be blended in with the main text, while corners might be chopped off, and in older printings where the inter-line spacing might not be exactly constant, whole lines can be elided. If I'm going to exert more effort in getting images and OCR, I've found that the place where it pays off the most is in previewing and correcting the recognition areas before letting the OCR do its work. -- RS