do-it yourself repro v-cradle for paper books

[user]

neither Casio? your exilims arent remotely controled from pc?

[ereszet]

No they aren't. But they fit in my pocket and have a long battery life. That's why I use them for my business trips, e.g. for photo scanning and e-mailing a signed agreement.

[user]

any impressions about new finereader?

I found it to be less faster and didn't see any significant improvement on accuracy in comparison with 8.0

[ereszet]

Originally Posted by user

any impressions about new finereader?

I found it to be less faster and didn't see any significant improvement on accuracy in comparison with 8.0

I must get used to it to give a fair verdict but my first impression of Finereader 9 is a big disappointment.

Djvu files are not supported neither as input nor as output.

Saving of blocks (text and images) is possible for the selected page only (I will still have to use my external script to get rid of black lines and blobs for all the pages).

Cropping (added hastily in version 8) has not improved. Small window, rectangle with 4 lines that have to be adjusted separately (can't they learn from Picasa or Irfanview? - just one mouse movement), batch cropping does not take into account odd or even pages, preview is limited only to the selected page.

Processing window (where current processing info is displayed) cannot be put to background, so the Finereader window cannot be reduced to an icon in Windows menu bar.

But the most annoying thing is that it doesn't work most of the time. Opening a file takes so long that I try to cancel the operation, which again takes so long that I use Ctrl-Alt-Del. Is it just me? 512 KB of memory and 2.8 GHz is not enough? I attach a screenshot of an error message telling me of a timeout regarding some server. Is Finereader trying to connect to some server (on the net?) that is not responding and that halts the operation?

I hope that in a week or so I will find time to contact Finereader support to get a feedback from them. But for now, I use my old Finereader 8.

[user]

Originally Posted by ereszet

Processing window (where current processing info is displayed) cannot be put to background, so the Finereader window cannot be reduced to an icon in Windows menu bar.

this can be easily done with an autohotkey script I suppose

Originally Posted by ereszet

(I will still have to use my external script to get rid of black lines and blobs for all the pages)

I d be glad if you would like to share any time-saving or process-improving scripts

Originally Posted by ereszet

But the most annoying thing is that it doesn't work most of the time. Opening a file takes so long that I try to cancel the operation, which again takes so long that I use Ctrl-Alt-Del. Is it just me? 512 KB of memory and 2.8 GHz is not enough? I attach a screenshot of an error message telling me of a timeout regarding some server. Is Finereader trying to connect to some server (on the net?) that is not responding and that halts the operation?

Originally Posted by ereszet

But the most annoying thing is that it doesn't work most of the time. Opening a file takes so long that I try to cancel the operation, which again takes so long that I use Ctrl-Alt-Del. Is it just me? 512 KB of memory and 2.8 GHz is not enough?

I agree. It takes disproportionately too much time. T2050 Dual Core and 2GB RAM here.

Originally Posted by ereszet

I attach a screenshot of an error message telling me of a timeout regarding some server. Is Finereader trying to connect to some server (on the net?) that is not responding and that halts the operation?

I think this is the license server that needs to run while finereader runs in trial mode (NetworkLicenseServer.exe)
maybe your firewall blocks it

[JSWolf]

It could be that Finereader 9 loads something that acts as a server, but is either not loading or not working.

[user]

Omnipage has introduced new technology for camera OCR

anyone tested it?

EDIT: I just checked it, don't waste your time, Finereader is still the best

[dbrotzen]

Originally Posted by ereszet

Minimum Requirements
. 2–megapixel sensor
. Variable focus lens (fixed–focus cameras, common in cell phones and hand–held devices, will usually produce images unsuitable for OCR)
Recommended Requirements
. 5–megapixel sensor
. Flash disable feature
. Manual aperture control or aperture priority mode
. Manual focusing
. An anti–shake system, otherwise the use of a tripod is recommended
. Optical zoom

Hello ereszet,

The info you provide is fantastic. I just discovered this thread. Thank you!

I travel a lot and imagine bringing along a pocketable digital cam to photo scan magazine articles, and later ocr them, enabling search. I contacted abbyy for any hardware recommendations beyond what you quote above, but they did not want to give any other advice. Previously I've used a C-pen to scan text. Slow and headings go amiss etc. Rather than being at my desk with a tripod and v-cradle, I imagine sitting aboard a plane, taking out the camera to free hand shoot a page of some text I'm reading, using the settings/features abbyy listed. I've got a few basic questions:

1. Capturing images with a handheld camera with image stabilization, (obviously the results wouldn't be nearly as good as with tripod and v-cradle) but do you think it could work? Have you tried hand held photo scanning and later OCR?

2. I wonder, would a wide-angle lens be a help or a detriment? I just noticed I had to be quite far away from the page when taking a picture from a book with a Sony DSC-200 I've got available and a wide angle lens would enable being closer to the page when taking the picture. I have not tried any OCR software.

Thank you for any feedback!

[mlbspike]

A lot of practical information here, about the DIY option. I've been looking into this for a while. I knew Atiz was coming out with a smaller unit, at a more reasonable price. nearly $1600 is not cheap, but for a pre-built solution, it looks attractive ... the previous version was not only huge, but was nearer $4500. ereszet ... As someone with realworld experience, with building your own, and with the issues you've had to overcome, I'd be interested in your opinion of the product (obviously, there's only so much you can tell from the ATIZ website, without actually having a unit). For myself, as much as I like a DIY project, my time and patience is in limited supply currently, and the pricetag has finally come in to a very acceptable range ... regardless of the laughter from some who will no doubt call it a ripoff.
- Mike

[Studio717]

mlbspike, if you can, please post your thoughts about the device once you receive it and have had a chance to use it. I'd like the set up, but I'm wary of spending that kind of money if a home-made set up will work almost as well.

I have one camera that I think would work fine with it, but in addition to the initial cost would be the cost of a second camera.

[Studio717]

ereszet, have you posted the plans anywhere for the hinged-side cradle? My apologies if you have and I missed the link.

[recycledelectron]

You've inspired me. I've bought a tripod, and plan to build a cradle over Christmas break (starting Dec 14, and running for 3 weeks -- I'm a college prof.)

Andy

[recycledelectron]

OK...I built my cradle, bought a tripod, set up some lights, and tried my copy of Abbyy FineReader 8.0. I've got a Samsung S630 (6.0MP point-&-shoot) camera.

I get good results except on the bottom 2 to 5 lines of text. They seem to curve / warp, even with glass on top, and Abbyy does not handle that well.

I tried Atiz's free software to correct distortion from photographed books, but it seems to only work with both pages in book mode. (I think you have to have used the S/W to know what I mean.) Also, the lines it draws around the page are often going to take the whole left page, and only half the right page. I can not adjust the right border of the right page. Has anyone else had this problem?

Andy (the anarchist)

[megalomania]

Greetings MobileRead community, and to ereszet in particular. I found this forum and thread via one of my periodic searches of the net for anything related to scanning books with digital cameras. For over 4 years now I have been digitizing my books on and off, and every few months I like to see if anyone else is doing the same thing. It is only within the past 1.5 years that this kind of information has started to proliferate on the Internet.

A little over a year ago I built my own V cradle based on a picture of the Atiz BookDrive DIY. It does not look as nice as the one ereszet built, but the principle is identical. I have images of the cradle and building instructions in a thread at my website: http://roguesci.org/theforum/showthread.php?t=1232 I actually posted about my book cradle starting on page 5 of the thread.

The important innovation of my cradle is a horizontally sliding base that moves the book every turn of the page to maintain the exact same distance from camera lens to paper surface. In close-up (macro) photography there is a very significant difference in the area photographed by moving an inch closer or farther away. Imagine a 1 inch thick book, where page 1 is the bottom of a triangle, and the camera is the peak of the triangle. Now picture an imaginary line of the triangle extending to the bottom of the book, to the last page. If your camera is very close, this new triangle will be 1-2 inches wider on the last page than on the first page. If you do not readjust your camera after each picture is taken, every time you turn a page the resulting surface of paper is father away from the lens, and the resulting text is smaller. The distance increase is only equal to the thickness of the book’s paper, but with hundreds of sheets of paper it adds up to a big difference.

For all practical purposes you can manually adjust your camera every 50-100 pages or so, but if you try to compile an ebook to look nice, you will discover that batch cropping the pages is difficult because you have slightly more margins to crop with every page. If you rapidly scroll down such a PDF document you can see the text get smaller, and smaller, and smaller. If you try to OCR the pages the problem really gets bad. Page 1 is perfectly centered and cropped, and it might be 305 dpi, but page 100 has wasted margins because it was much farther away, so it is only 260 dpi. That is quite a significant difference in resolution.

My original cradle used a drawer slide to move from side to side, but the sliding action is too jerky, it does not allow for the fine fraction of a mm movement of each turned page. I bought a linear slide, which is a rather expensive device that moves from side to side with minimal friction. Linear slides are made for precision fractional mm movements. There are usually a few being sold on ebay for around $50, which is a good deal considering the typically cost $200-$300 new from the manufacturers. Once I finish wiring up my lighting, I will finish the copy stand and camera mount, complete with clear glass platen. The final device should be a close approximation of planetary scanners like the BookDrive, but not for $4500.

On the link to the thread I gave above I believe I posted about calculating dpi from a digital camera. I am writing an ebook about camera scanning, so I have a considerably more thorough treatise on the meanings of DPI, PPI, megapixels, etc., and how it relates to a final image. In brief, if your digital camera has a paltry resolution of 1600 x 1200, which equals 1.92 pixels (would be labeled a 2 MP camera), each of the 1600 pixels corresponds to DPI. If you photograph a 1 inch wide piece of paper, you would have a resolution of 1600 DPI. This might work for a newspaper article, but books are not 1 inches wide, they typically range from 6 x 8 to 8 x 11 inches including margins. Since cameras have a ratio of 4 x 3 (1600/4 * 3 = 1200, hence 1600 x 1200), but books have a ratio somewhat less, you only need to consider the width part of the camera (1600 in my example) to photograph books sideways from top to bottom. An approximately 6 x 8 inch book with half inch margins can be zoomed in and close cropped to 7 inches, which still leaves a little safety room in the margins. Our 7 inch wide picture only has a resolution of 1600/7 = 228.6 DPI. This is good enough to read, not so good for OCR. If you fail to adjust your camera every so often and the later pages are farther away from the lens, say we are now digitizing the full 8 inches of the book by page 400, our resolution would be 1600/8 = 200 DPI.

Figuring out what resolution camera you need to scan books is entirely dependant on how big the book is. Indeed the more megapixels you have the higher your resolution will be when photographing the same area. However, digital cameras do not technically correspond exactly to DPI like a scanner does. The most important factor of a digital camera is the lens. The lens makes or breaks the book. When I first started digitizing books I bought a 6 MP handheld point-and-shoot camera that had settings for close-up shooting. I scanned a few books for a local bookstore that deals with very rare 19th century local history and genealogical books. Demand for these books was high because people wanted to look up their family history, but there were few of these books ever published, and the condition of the books does not allow them to being manhandled. Flatbed scanning was out of the question because the books would be damaged by laying them flat and pressing down.

A bit more about old books, overhead scanning by a digital camera is the best solution for books in bad condition, or that are very valuable. It goes without saying you can not razor valuable books and scan them in a sheetfed scanner, and getting a bound book flat on a flatbed scanner requires a great deal of spine crushing pressure. Books with a damaged spine should never be opened more than 100 degrees. As I was designing my book cradle I found professionally manufactured book cradles for libraries invariably have 90-100 degree angles. At 90-100 degrees the pages on one side of the book will be absolutely flat. Open a book and see for yourself; if it’s a hardbound book or thick paperback, open the left side to page 20 and the right side will be flat until you open the book past 90 degrees.

I am digressing from my point about lenses. My little digital camera did a very good job of digitizing the pages, until I went to OCR the pages. Much to my dismay I discovered the corners of nearly all the pictures were slightly blurry. As it turns out this is called spherical aberration, and is a common defect of cheap lenses. Consumer model handheld cameras are designed for casual portraits at distances from 5 feet and greater; a zoom lens is not made for precise corner to corner focusing because it is a general use component. These lenses are a Jack of all trades, master of none. My camera, and any consumer camera, is unable to keep both the center area and corners focused at the same time.

My solution was to move the camera back farther to increase the amount of wasted margin space, but this also lowers the effective DPI. Then I bit the bullet and bought a professional DSLR camera with a detachable lens, specifically a Rebel XT with a macro lens. Macro lenses are designed for close up corner to corner focusing. To put the importance of the lens into perspective, my macro lens costs more than your entire handheld digital camera. Even a low quality lens for a SLR camera costs as much as cheap digital cameras. The lenses on any handheld camera are only realistically able to utilize a maximum of 10 megapixels. I recommend everyone read about the megapixel myth before buying one of those 12+ megapixel camera. Without a high quality lens, as with SLR cameras, imaging sensors beyond 10 megapixels are just useless marketing fluff that trick consumers into thinking “my camera is better than your camera.”

Commercial units like the Atiz BookDrive and their new BookSnap, and industrial models like the Kirtas book scanners all use quality cameras with quality lenses. The BookDrive uses Rebel XT and XTi cameras with macro lenses as a matter of fact. This does not mean your little digital camera is useless, far from it, just don’t expect perfect ebooks just as good as the publisher can make. My top priority in digitizing books is OCR; I digitize scientific reference works so they can be searched for the rare snippet. I rather dislike reading ebooks of any kind, on any reader, I would much rather have a paper copy. I realize this position may not be popular with the members of this website. If human readability is your top priority, and OCR accuracy is only secondary, than a handheld digital camera is more than adequate.

In my early experiments with digital cameras I found 3 megapixels to be the minimum resolution to digitize 6 x 8 inch books and still be able to read the text. Because the lenses are so bad on low resolution cameras, like camera phones, digitizing a smaller area can still be difficult to read. According to my calculations I will need a 70 megapixel camera to do 600 DPI scans, two pages at once, of 8 x 11 books. I won’t be holding my breath for one of those anytime soon, not an affordable camera anyway.

I can go on and on about methods to digitize books with a camera, but I will spare everyone further reading. I reserve the numerous fine details for the ebook I am writing on the subject. It will be free and not copyrighted, and I suppose I will include an online version on my website as well. Taking the pictures is actually the easy part, post processing the images is the challenge. Post processing has to be fast and automated, otherwise you lose the time saving advantage using a digital camera gives you.

Incidentally, I read in another thread here, https://www.mobileread.com/forums/showthread.php?t=14475 , that some people were questioning why using a digital camera is a superior method. Overhead scanning is absolutely the single fastest method of scanning a book. I hate scanning books, it is the most boring, tedious, mind numbing task ever devised. Being able to digitize books in less time was just the motivator I needed to actually start scanning. A digital camera scans a page in a fraction of a second, compared to the 45-60 seconds of a scanner, or 3-4 seconds of a copy machine. However, there is more to scanning a page than just acquiring the image. On either a copy machine or flatbed scanner you have to flip the book over, turn the page, flip it over again, align the book on the glass, and then scan the page. This adds another 20-30 seconds per page, maybe more if you are not so agile. With a digital camera I can scan a book almost as fast as I can turn the pages. It takes me an average of 5 seconds to lift my glass cover, turn the page, lower the glass, and press the shutter release on my remote control. I will put a movie on and get into the zone, just turn the page and click. Using the camera is such a passive activity that it can become an easy routine, and I like to believe I am not wasting my life watching TV if I am scanning books at the same time. The constant flipping and aligning of books on a flatbed scanner requires more attention to what you are doing, and this is more like work. On a copy machine I doubt you can have the luxury of TV or music to distract you.

[ereszet]

Originally Posted by megalomania

A little over a year ago I built my own V cradle based on a picture of the Atiz BookDrive DIY. It does not look as nice as the one ereszet built, but the principle is identical. I have images of the cradle and building instructions in a thread at my website: http://roguesci.org/theforum/showthread.php?t=1232 I actually posted about my book cradle starting on page 5 of the thread.

The important innovation of my cradle is a horizontally sliding base that moves the book every turn of the page to maintain the exact same distance from camera lens to paper surface.

Your experience and experiments have been very similar to mine. I am quite satisfied with the results I get and document/book scanning has been a routine thing for me for more than a few years now.

I have visited your thread at the site you indicated and looked at the sliding base photos but I cannot understand how the base moves the book at every turn of the page. I guess how Atiz BookDrive does it (cradle moves horizontally while the transparent cover moves only vertically) but I cannot see any cover mechanism in your solution.

As for my experience, I am using PSRemote program to control the operation of my camera from my laptop, so I can see the picture in advance on the screen and I can correct the position of the cradle a few millimiters every 20 or so book pages scanned (no sliding mechanism required).