DIY book scanner, 150 pages/min

[Turtle91]

Originally Posted by BeccaPrice

With 1dollarscan, the pdf is OCR'd. What I did was save the PDF as a TXT file (I"ve got Acrobat Pro, so I can do that), and then had a file I could edit for scanning errors.

I was under the impression that Acrobat - even Pro - doesn't keep the formatting when you save to text. In which case you will not have any of the italics, bold, superscript, etc.

Assuming the PDF they give you is a perfect OCR of the original - you would still need to go back and manually format the entire book to make it like the original.

I did an experiment by creating a test page in Word with different formatting of sections of text. I then saved that document as a PDF. This provides a "perfect OCR of the original image". When I opened that PDF in Acrobat Pro, everything looked as it should and I could perform a find on any of the words in there. I then saved the PDF as text. Acrobat gives 2 options, Plain text and Accessible text - I did both. In both cases the text was correct but without ANY formatting.

If there is a different way of saving a PDF to text, I would be very interested to know how.

Sample OCR text.pdf
Sample OCR text - plain.txt
Sample OCR text - accessible.txt

[rkomar]

Originally Posted by Turtle91

No. It will scan the text and OCR, but computers aren't smart enough to know what it MEANS...at least that I've seen.

If you wanted active hyperlinks in an ePub/html you would need to manually insert the links to the proper locations. If all you want is a digital copy you could save it in PDF.

I scanned my old textbooks and references to free up shelf space. Because they were so technical (equations, tables, code snippets,...), I couldn't really OCR them and turn them into epubs. So, I saved the cleaned up images as PDFs, and added a table of contents to each using PDFMARKs (those bookmarks you see on the left in most PDF readers). It's a fair bit of work, but I think it's necessary for a reference to be useable. I didn't do it myself, but you can also OCR the text and add it as an invisible layer in the document to make it searchable. Of course, you can only search the text that OCRs perfectly.

[AnemicOak]

Originally Posted by Turtle91

If there is a different way of saving a PDF to text, I would be very interested to know how.

With Acrobat Pro I usually save as HTML or RTF, which usually allows formatting like italics to be kept.

[Turtle91]

Originally Posted by AnemicOak

With Acrobat Pro I usually save as HTML or RTF, which usually allows formatting like italics to be kept.

I hadn't used Acrobat's save to HTML in a while. Previous versions weren't very good - thus part of the "nightmare of converting". But I have a new(er) version (10) and tried it out.

It is fairly clean...better than before...and you are right it saves bold and italics...but it still has some issues. On this very simple test page there are several formatting discrepancies that would need to be fixed...not impossible with search and replace, but very time consuming. I would be hesitant to try anything more complex or longer than a simple page or two.

Thanks!

Sample OCR text.html

[AnemicOak]

The output from Acrobat Pro can vary greatly depending on the original source and what tools were used to make the PDF (and possibly the PDF version). I hate converting PDF, but sometimes it's the only source and I usually find the HTML export to be the lesser of evils so to speak. On a few occasions I've gotten better HTML by importing the PDF into Mobipocket Creator.

[Iznogood]

I actually have built myself one of these scanners - as a Christmas present for myself - and it works fine. As a student at the technical university in Norway I have access to a CNC router and could make the parts quite cheaply. With a hi-res camera, I can scan ~600dpi in color in about 4 seconds. It takes so long because I store the captured JPG-image and the raw-data directly from the camera, pluss I process the images in real-time instead of post-processing. If I only had taken the compressed jpg-image, it would have taken about a second (max two seconds) per dual-page.

Right now, I'm getting acceptable results. I have some problems with reflections in the glass, but that seems to be unavoidable with this design. I am experimenting with techniques for removing reflections and glare, but have not been 100% successful.

Rotating, descewing and OCRing can be done quite efficiently as post-processing by quite simple scripts. All in all, this is a very fine scanner for books. It can scan my old, valuable books without destroying them, and I now can read them (i.e. the digital copies) in bed while the originals slumbers safely in my shelves.

[kevinp]

Originally Posted by Turtle91

I was under the impression that Acrobat - even Pro - doesn't keep the formatting when you save to text. In which case you will not have any of the italics, bold, superscript, etc.

Assuming the PDF they give you is a perfect OCR of the original - you would still need to go back and manually format the entire book to make it like the original.

I did an experiment by creating a test page in Word with different formatting of sections of text. I then saved that document as a PDF. This provides a "perfect OCR of the original image". When I opened that PDF in Acrobat Pro, everything looked as it should and I could perform a find on any of the words in there. I then saved the PDF as text. Acrobat gives 2 options, Plain text and Accessible text - I did both. In both cases the text was correct but without ANY formatting.

If there is a different way of saving a PDF to text, I would be very interested to know how.

The (so-called) OCR in Acrobat is mainly just so you can search for text in the PDF. It's not made to do what you are thinking.

That's why you need to use an actual OCR program. I use Abbyy. It will open a PDF and extract the pages as TIF files, then do it's thing. And it works fairly well on stuff like paperbacks. It will capture bold, italics, etc.

If you want to OCR stuff like textbooks that contain lots of illustrations and such, I don't know of anything that works 100%.

[Kirtai]

Originally Posted by HarryT

Libraries? It's one thing to scan your own books - quite another to scan library books. That's just blatant copyright infringement.

Actually, you are allowed to copy parts of many kinds of reference documents at a library. For instance you're allowed to copy up to 10% of an official British Standard.

[danielreetz]

Originally Posted by Turtle91

Hey Daniel!
I was wondering when your ears would start twitching and you would notice this thread!

It's good to hear some actual numbers instead of relying on my old memory.

I hope I got most of the info right.

Cheers,

You sure did, thanks Turtle. I thought I recognized your username.

[danielreetz]

Originally Posted by Iznogood

Right now, I'm getting acceptable results. I have some problems with reflections in the glass, but that seems to be unavoidable with this design. I am experimenting with techniques for removing reflections and glare, but have not been 100% successful.

I think we're talking about this on the DIY Book Scanner forums, but have you tried moving your lights further up from the glass? Glad to hear you were able to get your machine together!