From physical to digital
 

I have a somewhat large library of real books. They're a real PITA to carry around, and even worse to ship for a household move. So, I'd like to consider scanning in my book collection for use on an ereader. Some questions:

- Do most people just scan to .jpg and format for the screen?

- If so, about how large of an image file is each page?

- What tools do you use to batch process the files?

- auto crop and rendering to for a specific device's resolution


Or are folks using OCR to convert to text?

- Do you notice significant errors creeping into the text?

- What do you do to fix those errors?

- How do you find obvious text errors without manual editing?

BTW: I just bought a Sony PRS 505, but I expect to buy one of the larger full page units as soon as they hit the market (or the wireless iRex unit with a functional firmware hits).

Don't know about most people, but I OCR the scans and take pains to proofread and correct them.

In 300DPI PNG, which I use, something between 1.5 and 3 MB per page, depending on complexity. Much less with JPEG, obviously.

Way too many, yes.

They need to be hand-fixed.

A few typical errors can be found and fixed with regular expressions, but mosty of them require manual approach.

Guys, a method I've found to improve the quality of the OCR process involves photocopying of the book's pages first, expanding the page image up to letter/A4 size. Then scan those letter-sized pages into an OCR scanner. Many of the better OCR scanners can allow standard-sized paper to be fed into them and read at high-speed, removing the need to manually scan each page (though you'll still end up doing that at the earlier copier stage). And the expanded letters will be easier for the OCR program to read, resulting in fewer errors.

Personally, I feel the 2-step photocopy-scan process is worth the creation of scanned pages with fewer errors.

Occasionally, you luck out and discover that a particular error happens regularly, and you can fix it with a "find-and-replace all" process. But you should still go through every page manually.

You could just as well scan in higher resolution, e.g. 600 DPI.

Actually, the scanning process is a lot less demanding than I originally thought. With Plustek OpticBook 3600, I am doing slightly more than 100 pages every 20 minutes, getting a normal-sized book completely scanned in about an hour.

Do you cut the book up with a razor and then feed the pages through the feeder? I have a Brother multifunction unit with a feeder that supports double-sided scanning. Or are you doing everything possible to save the original copy and binding?

With OpticBook, there is no need to damage the book.

Professional autofeed scanners (the kind used by document printers like the Xerox DocuTech and similar platforms, found in many on-demand print shops like Kinko's) can scan 100 pages in 2 minutes or less, and with the right OCR software, generate scanned images of those in under 10 minutes. (Not that the Optibook rate is bad, I'm just saying there are faster methods that also work well.)

Usually, the only catch is getting shops that have this equipment to allow you to use it, as they tend to assume your scanning a book to do an OCR is probably illegal...

They are, but
1) I can't afford the non-destructive ones, and
2) I don't want to use the destructive ones.

Right. But remember, I said to photocopy the pages first, which does not have to be destructive... and run the photocopied pages through the scanner. Trust me, I have done this, and it's not so bad, and not that hard on the book (provided it doesn't have a flimsy spine).

I'm torn as to which method to use to convert some books so that I can read them on my Sony 505. Should I plunk down the cash for a Opticbook scanner or rig up a homebrew setup to take pictures of the pages? I want to save money but I also want the fastest method. I was contemplating getting a new camera soon anyways?
What is the width of the lip on the Opticbook scanner anyways?

Not sure what you mean by "lip", but I assume it to mean "how much space do I need between the spine and the start of the text". With OpticBook, you need at least 6 milimeters or so, with 8-10 milimeters being comfortable enough. It depends on the tightness of the book, of course, and I am having much easier time with hardcovers than with paperbacks.

Do you find that you have to hold the book down to get it to stop curling up? Is that why you say that hardcovers work better. How does the Opticbook scanner work with Abbyy Finereader? How long would you say it takes to scan say 100 pages or so?

Thanks!

I probably wouldn't have to hold it, but I get better results if I do.

Mostly because hardcovers are a lot more tolerant to full opening without getting damaged. Paperbacks, I tend to read while opened less than 90 degrees to preserve the spine.

I have never tried scanning from inside FineReader (*), but when I scanned in the application provided with the scanner and then imported the images into FineReader, it worked just fine.

*) The scanner and its software were apparently created by someone who thought long and hard about how it would be easiest to scan books, and it shows - it really is very comfortable and easy to do. Most of my scanning is done without looking at the screen, with just occasional glances to make sure everything is still fine.

It really depends on the book, but I average 100 pages every 20 minutes (give or take a minute).

It's a matter of which kind of binding they have. Hardcovers are usually sewn, while paperbacks are just glued (or whatever the right terms are).