Materials: hardwood, glass, steel rods, nuts and bolts, black spray, black cardboard.
Cost (materials and work by a friendly handyman): less than $50
The attached two pictures say it all.

Part 1 was in my post Do-it-yourself repro v-cradle

1. Why a v-cradle and not a flat bed.
For two reasons: First, you cannot spread the books flat and if you do, the quality of reproduced pages will be compromised. It is especially important if you plan to OCR the book. Second, you avoid light reflections. You need only one lamp with a diffusor just over the v-cradle (picture attached). Lighting is the most difficult part of reproduction. Over the years, I have tried various setups with my semi-profesional Manfrotto repro stand and four lamps at 45 degree angle. It doesn't come close to an overhead lamp and v-cradle. Avoid any other light in the room or take everythig to your terrace and shoot at the sunlight with no artificial light.

2. Where to put the camera
On a tripod, at 90 degree angle to the book page. The camera and the tripod should be masked with a sheet of black cardboard to avoid reflections (picture attached). The best camera is black camera anyway. Same goes about tripod. I use now 8 and 10 megapixel cameras, but 5 megapixels are good enough for most A4 documents.

3. How to shoot fast

Quick and dirty: from hand (or tripod plus remote) if somebody turns the pages for you and keeps them flat (fingers may be visible at margins but you can erase them or the margins later). Takes less than 5 seconds per page.

Precise: from your laptop connected to the camera (My Canon Powershot Pro1 makes it easy - a picture attached). The camera can shoot automatically in regular intervals, so you need no help from your buddy. Takes up to 15 seconds to flatten the page with a glass weight, re-position the cradle if necessary, and shoot. About the same time is required per photo to download (USB 1.1) in a batch after the shooting session. Before shooting use a test page to fix the focus of the camera. Re-positioning of the cradle (rather than the camera), i.e. moving it closer to or further from the tripod, is necessary after shooting 20 or so pages if you do not want the opposite page to get into the picture (a viewfinder on a laptop screen is much better for positioning than using the one on the camera). Alternatively, you can place a black cardboard over the opposite page. Removing black margins aftewards is easy with proper software.

More to follow, if there is any interest.

Camera setup: view from the front and the back.
Light diffusor.

That's a very nice system you've got there, I suppose that if you had 2 cameras, and your software could handle it, you could shoot both pages without moving much.

This looks like a pretty viable alternative to the >$10k automatic system that someone posted about a year ago. I'm certainly a lot more likely to do something like this than shell out for the library-grade machine. :nice:

The book holder looks like it's the most complicated part of the assembly -- do you find that a particular angle is best for the shots? I'm wondering how much you'd lose by going with a fixed angle holding bed, you see. :chinscratch:

Thanks for posting these, ereszet. :yes:

That's a very nice system you've got there, I suppose that if you had 2 cameras, and your software could handle it, you could shoot both pages without moving much.

This looks like a pretty viable alternative to the >$10k automatic system that someone posted about a year ago. I'm certainly a lot more likely to do something like this than shell out for the library-grade machine. :nice:

The book holder looks like it's the most complicated part of the assembly -- do you find that a particular angle is best for the shots? I'm wondering how much you'd lose by going with a fixed angle holding bed, you see. :chinscratch:

Thanks for posting these, ereszet. :yes:

The system with 2 cameras (company name is Atiz) looks very nice and costs now much less but sitll a few thousand USD plus cameras. It is semi-automatic - you turn the pages by hand, but v-shaped glass/plexi? can be moved up and down without removing it, and the shots are taken automatically.

Apart from a v-shaped glass and a mechanism to raise it, my system does exactly the same plus a regulated angle, distance, and room for books of different thickness. Two cameras are no problem technically, they can both shoot automatically every nn seconds (at least my Canon can). The problem is quality. You should concentrate on making every shot picture perfect. Otherwise there is more manual work at the processing stage. Now I shoot all the odd pages and then all even pages just by turning pages and adjusting the distance from the tripod by a few milimeters every 20 or so pages.

The book holder is just two panels with two hinged supports plus a heavy base with a lot of grooves. This is my patent :) You can regulate the distance between two panels, the angle of each panel, and the distance between the cradle and the tripod. The regulated v design is crucial to ensuring the angle between the lens and the book page at 90 degrees plus avoiding any skewing. Precise correction of angle and skewing with the tripod controls is very cumbersome. Since I installed my system, I do not touch my tripod any more (I miss my 7x zoom when taking photos of my garden birds - like the barbet in my avatar - with other 3x zoom cameras).

As for cheap and dirty alternative you can make something similar from a cardboard. It was my first design. My wife used her scissors to make it in one hour or so. It worked perfectly. Remember, the devil is in the lighting and reflections, especially for color and glossy pages (or pages under glass). The disadvantages of cheap and dirty solution are: it is very light and therefore unstable, and you have to change tripod postion controls frequently to accomodate for various sizes of original pages, since moving the cradle further away creates a trapezoidal effect. You can avoid it with changing the angle in the v-cradle. The focus is quite forgiving in such a case (+/- a few milimeters difference in distance between the lens and upper and lower parts of the page do not show in repro).
The advantage of the cardboard cradle is portability.
Finally, you can always use the simplest solution: a friend or a document holder (picture attached) and soft flash plus a lot of processing to get pictures straight. You need a steady hand and a camera with a good automatic focus (out of my two Casios, ex-z850 does a better job than ex-z1000).

Thanks for asking NatCh.

The angle between the lamp and the book page can be anything (usually 45 degrees) as long as the reflected at 90 degrees light does not go into the camera lens.
For uniformity of lighting the best thing is natural light outside, an apartment corner with two windows at 90 degrees or a bright lamp as high as possible (a bright ceiling lamp will do) over the cradle plus a diffusor. The v-cradle assures the proper angle.
Typical setting of two or four repro lamps at 45 degrees to the book page may work with very expensive lamps that I cannot find even in best photo shops (however there are some ads on the net). But in my experience, even if you see no bright spots on the page, the reproduction will show them. Human eyes provide for automatic correction, while the camera is not forgiving at all (photographers who deal with RAW pictures know it too well).

Cradle distance and angles

First picture to the left is a setup with camera lens at 90 degrees to the book page placed in the v-cradle.

The middle picture shows that moving the v-cradle closer to the lens helps to maintain 90 degrees angle for pages of bigger sizes. The zoom in the camera has to be adjusted accordingly.

The last picture shows what happens when the distance between the cradle and the lens stays the same despite a bigger page size. The distances between upper and lower parts of the page and the lens differ considerably, which causes a trapezoidal effect.

Not shown in the pictures but an obvious solution for bigger pages is to change the lens angle in order to keep the focus in the middle of the page. But when the lens angle changes then it is no longer 90 degrees to the page. That is why you can change the inclination of v-cradle panels.

Cost: nothing (just a diamond ring for my wife, but I would buy it anyway)

Enclosed pictures are worth a thousand words (and bucks if you had considered to buy an expensive substitute :) ).

Cost: nothing (just a diamond ring for my wife, but I would buy it anyway)

Enclosed pictures are worth a thousand words (and bucks if you had considered to buy an expensive substitute :) ).

This one looks doable, even for me. Does anyone have OCR software recommendations that work with camera images?

Phrodod

This one looks doable, even for me. Does anyone have OCR software recommendations that work with camera images?

Phrodod

Finereader 8
Takes images with resolution from 96 dpi upwards (I tried 72 dpi with success). Ingenious recognition algorithm with all imaginable languages. Splits double pages. Converts to black and white. Can save blocks of text and images only (removing any useless white or other space around - requires some trick though to do it in batch). Finally, saves as text under picture pdf format.

IMHO The best available OCR software. I have bought it. I put my money where my mouth is.

BTW I am considering posting some hints on processing photos to a book format but I am not sure if this is a right forum/thread. Any ideas?

That's a very nice system but do you use special software to handle it.
Can you give us hints to make the best results.

That's a very nice system but do you use special software to handle it.
Can you give us hints to make the best results.

You either take time to position the book right in the cradle (no distortion, no background behind the page, uniform lighting - just rectangular page in your viewfinder) and then no processing is required apart from conversion of images to pdf or djvu with any freely available software. Or you shoot fast with the black background and don't care about photo quality, which you improve later with proper software. I prefer shooting fast because processing later is fun.

In both cases free Google Picasa can be very useful (downloading the pictures from the camera, automatic contrast and colour corection in batch processing, individual deskewing and cropping, etc.). Free Irfanview may be useful as well, especially for batch conversion between image formats.

I will post my suggestions for more sophisticated photo processing software (for low quality photos like the ones offered by google books in pdf format or a number of digital libraries in djvu format - you know dLibra, don't you) over the weekend.

Before photoscanning hundreds of pages, it is worth making sure that the lighting is uniform. Neither your visual inspection of the original paper document nor the camera LCD display or the computer screen (if you shoot from a camera connected to a computer) are good enough to see the bright and shadow areas due to non-uniform lighting (unless shadows and bright spots are very discernible).

There is a commercial program (Psremote) that in its preview mode can highlight the areas that are too bright according to a selected level of luminosity. However its use is limited to cameras that can shoot from the computer and it can only show which areas are too bright at a chosen threshold rather then areas with different levels of brightness.

I deviced a method which is simple and effective to check the lighting uniformity before shooting hundreds of photos.

You just take a test photo as jpg and compress it aggresively using Irfanview or any other similar progam. On the compression scale of 100 (best quality) to 1 (lowest quality) you may choose 10 or lower quality. The resulting compressed jpg will show "clouds" or areas that are similar for the sake of compression. For our purpose, it shows areas of pixels with similar luminosiity. By forcing a very low quality, we can see "clouds" even in a photo that is otherwise almost perfect. Once we know which parts of the test photo are under- or overlighted we can: move the cradle, move the lamp, use a curtain or a black bristol paper to block the light from reaching the original at various points and angles.

The attached pictures show:
Daylight.jpg - a test photo taken in the middle of the room with two windows and various objects in the way of light (the gray line along the photo image is the feature of the original paper page).
Daylight10.jpg - same photo compressed at the level of 10, showing the clouds of non-uniform brightness
DayTungsten.jpg - like daylight.jpg but with a ceiling and overhead lamps switched on (I do not recommend this setup, but I have no way to stop sunlight during the daytime - blinds are too transparent. Usually i shoot at night.
DayTungsten10.jpg - same photo compressed at the level of 10

You wll notice that the better quality DayTungsten.jpg shows some non-uniformity. It is due first of all to the sunlight but also to the fact that upper part of v-cadle panel is closer to the overhead lamp than the lower part of the panel. That is why I recommend using a bright ceiling lamp only. At at distance of 1.5 or 2 meters (depending on how high is your ceiling), the 20 centimeters between the upper and lower part of the cradle panels make little difference. You can also put a white paper (you can experiment with its dimensions) on the panel opposing the original page, and it will act as a kind of mirror, directing more light to the lower parts of the original. One method of achievieng quite uniform light is using the camera flash directed to ceiling. The light reflected from the ceiling is more dissipated. I even experimented with double reflection, i.e. flash directed to --> a sheet of bristol paper (or a sheet of matt aluminium) --> ceiling --> v-cradle.

If you are a little ovewhelmed with all this, remember that one ceiling lamp will usually be ok. The brighter is the lamp, the easier it is to get the white balance (i.e. to make the white pages white rather than gray, yellowish or pinky, while keeping the print black) in the processing afterwards. I am not going into RAW processing, white balance/ISO/camera setting problems or commercial heavy weight and costly software to process your photos. I concentrate on fast shooting and batch processing. But if you plan to convert all your family photos to digital ones, you may want to learn about RAW image format and all the tricks of professional photographers. There is abundance of advice on that on the internet but rather little of what I give you here in terms of convenient fast practical arrangements.

Please note that the paper originals were put under the glass but the photos show no reflections thanks to my v-cradle design. Better results can be achieved with proper lighting, zoom, and processing (see Tungsten processed.jpg).

The question is: do we need to make the lighting perfectly uniform? My answer is:
yes, when we deal with color originals;
rather yes, when we deal with gray photos;
not necessarily when we convert the pictures to mono.

Unless you are a perfectionist and the original is in blue dark, my setup is very forgiving even with bad lighting.

BTW. There are some commercial programs that can process the photos and help you make the lighting more uniform, but in my opinion they require fine tuning for each and every photo individually. A proper test and light setting before taking pictures is the only way to avoid disappointment in the final stage of book processing.

How it's made?
It is my wife's plastic compact powder case with mirror. I made a hole in the cover opposing the mirror to hang the deflector on the camera (I used a red hot nail to make the hole). If the light reflected by mirror and ceiling is not uniform enough even with soft flash, I cover it (the mirror) with a small piece of paper or a piece of matt aluminium, which dissipates the light even more.

Does it work?
Yes, for black and white book pages or pages with light colors (even glossy). For pages in dark colors, the camera will cast a barely visible (but annoying - if you are a perfectionist) reflection of itself on the original paper page in the v-cradle (the camera faces the original page at 90 degrees). That's why I usually obscure the camera with large black paperboard, which of course prevents the use of camera flash, and I stay with my proven overhead lamp (which is 45 degrees to the original book page and 45 degrees to the camera lens).
The advantage of compact flash is portability. You can use it anywhere together with a tripod and my "quick and dirty" portable cardboard v-cradle (see my earlier posts).

Cost: a gift to your lady, a Dior powder case perhaps?

For software tools to process photoscanned images see post #177 (Software tools to convert paper documents to lrf with thanks to cacapee for pdflrf) and the following discussion in "Yet another PDF to LRF converter" thread in Reader Content forum.
To find post #177 use the linear method for browsing.

To get a link to a specific post, click the #NNN link at top right of it:
http://www.mobileread.com/forums/showpost.php?p=99716&postcount=177

ereszet;

Thanks for the posts on the Cradle, Cardboard Cradle, Detailed Instructions, Masking Instructions, OCR Recommendations, Lighting Perfection Tips, and Software Tools!

This is quite a huge, detailed, beautiful set of instructions. Thanks!

Any chance you want to take it the next step and add it to the wiki?

You've provided an incredible tutorial for the rest of us.

:2thumbsup:2thumbsup:2thumbsup:2thumbsup:2thumbsup

Phrodod

ereszet;

Thanks for the posts on the Cradle, Cardboard Cradle, Detailed Instructions, Masking Instructions, OCR Recommendations, Lighting Perfection Tips, and Software Tools!

This is quite a huge, detailed, beautiful set of instructions. Thanks!

Any chance you want to take it the next step and add it to the wiki?

You've provided an incredible tutorial for the rest of us.

:2thumbsup:2thumbsup:2thumbsup:2thumbsup:2thumbsup

Phrodod

I am glad to here the good words from you and I welcome your suggestion to produce a wiki tutorial. However, I am afraid that I cannot devote enough time for that in the near future. If you or anybody else would like to use my posts to produce such a draft tutorial (e.g. From paper to reader: hardware and software tools), I would be only to happy to contribute to its next version. In the meantime, I will post only a few more hints on the hardware setup in this thread (after all, I see no more problems to solve unless somebody asks for advice), and more on preprocessing software in the cacapee pdflrf thread forum.

Please note that I am neither a professional photographer, nor a reprographer, nor a software developer but I try my hand in a number of fields beyond my call of duty, which covers a lot of areas. It is to be more efficient in my professional life (and pastime as well) that I follow or sometimes precede new developments.

1. Do-it yourself - see http://www.frontiernet.net/~rjacob/copystnd.htm. The picture I attach here is from that page.

2. Semi-professional copy stand by Manfrotto (Bogen in the US).
I guess that this is something similar to what some US university libraries offer for the use of students (It's been quite a few years since I last time lectured in the US, so I am not current on that).

I bought the Manfrotto in Italy a couple of years ago together with lots of accessories:
micro ball head, a mini laser level (that I attached to camera flash rail to aim the lens at the middle of the document page), extra 500 watt halogen lamps, reflector lamps, a variable friction arm, micrometric plate, various brackets, quick change plate adapter, multiclips (to hold black sheets protecting from side light), etc.

I was quite proud of the results achieved with that setting, although I was never happy with the lighting and shadows.

Since I am now a few thousand miles away from home, I had to device something provisional. So I designed and constructed a repro v-cradle as a temporary solution. It proved better than the flat bed copy stand for at least four reasons:
- it is ideal for books, magazines, etc. because of the V shape;
- it allows to position the original paper document at any distance and angle to the lens without touching the tripod or other camera stand;
- it solves the problem of lighting (see desperate solutions for flatbeds in the attached pictures taken from the internet);
- it does not require a tripod or a stand with precise positioning, anything that holds the camera steady at about 45 degree angle to the horizon level will do.

3. Professional reprographic systems. There are dozens of them with sky high prices. They may be right for automated reproduction of thousands of pages per day, but not for quality of each and every page. I assume that google books uses something like that and the results are sometimes disastrous (see yourself at google books).

I believe that with my repro v-cradle setup, I can match quality of any professional reproduction and I can photoscan hundreds if not thousands book pages a day (individual sheets of documents are just too easy to photocopy to mention).

Thank you for this! I have an old copy stand stashed away. Definitely time to pull it out. :D

Shooting photos and processing them takes time but it is not necessarily my time, I let most of the burden to my computers.

I work in my home studio at a desk with my old laptop plus an infrared external keyboard (I have to get a radio keyboard one day) and a Linksys modem/router. In another room, I have a desktop computer with dual core processor. The computers are connected as a LAN through an Ethernet cable but wireless works as well (I use wireless to surf with Archos 704 wifi when I am in bed). I use Windows terminal services to operate the desktop computer from my laptop.

To my right, on another desk is my photoscanning setup: a tripod with a camera (connected to the laptop) and the v-cradle. I can turn document pages in the cradle without moving from my chair and I shoot photos with my mouse. The shooting is the only time when I need to give my full attention. After that I switch to the other computer and start the download of pictures from the camera to the computer. It takes time, but in the meantime I work with my laptop doing other things. Next, I run all the picture processing programs (like OCR, conversion to pdf, and pdflrf) on the desktop computer. It does not take my time at all. I work with my laptop, and only occassionally I switch my screen to see what is going on with the picture processing.

I do not mention all the external hard disks, a maze of cables, power supplies, UPS towers, US, various European and other adapters for different plugs, usb extensions, card readers, etc., plus a big fan to cool my laptop.

Somehow it all works most of the time. Every evening for half an hour or so I photoscan the documents that I want to have in my computer, then I OCR and index them. I have not yet met anybody else who runs that kind of paperless office (now with the Sony Reader and Archos 704 extensions). But I assure you that it works (until hard disk crashes).

Can you be a bit more specific about the actual image-taking stage? Do you use the glass on each page?

Do you:

- place the glass on the page to be photographed
- focus the camera to the right or left page (I realize this is done at the beginning, but not continuously)
- assuming all the computer stuff is set up, take the photograph
- lift the glass, turn the page, replace the glass
- take another image of the left or right page
- continue until all left or right pages have been photographed, then repeat the process for the opposite pages
- clean up images in software
- collate images
- save as pdf (or whatever)

Is this how you're doing it?

I'm asking because I thought I understood, but when I went to step through the process, I realized how much I didn't understand.

Thanks for sharing this. I have a large (quarto-sized, I think) book that is very heavy for a traditional scanner (both for the scanner and for me!), and your method looks like it will help me a lot.

Can you be a bit more specific about the actual image-taking stage? Do you use the glass on each page?

The heavy glass serves to flatten the pages. It is five millimeters thick. An anti-glare glass would be nice, but I use just ordinary glass and it is ok with my lighting arrangement.

For single sheets of paper documents it is not necessary to use the glass unless the sheets are creased, like the ones that come folded in the envelopes.

In case of multipage documents, magazines, and books, I manually set the focus for the first page (i.e. I let the camera to focus automatically and if I am happy with the result, I fix it for the rest of the pages). The automatic focus for all the pages can work as well but some of the pages can have no print in the middle, where camera focuses, and the image will be blurred. If the first or cover page is like that and the camera cannot get the focus right, I use any printed page to put it over the book in the cradle to set and fix the focus before photoscanning the book.

My Canon connected to a computer allows me to zoom, set the focus, and change all kind of camera settings without touching the camera itself, with the results seen in a compter window (in a viewfinder, preview or downloaded image window).

For my other compact cameras (both Casio - 8 and 10 megapixels), which cannot shoot from the computer, I use a universal remote trigger that can be attached to the camera (I bought it in a photo accessories shop). I let the camera set the focus automatically for all the pages, however for the pages that have nothing to focus on in their middle, I put a half sheet of printed paper over the blank part to make the focus possible. The image has to be later cleared of that manually at the processing stage.

As for the zoom, it is a good idea to set it to such an extent that the whole page of the book (with no background visible) is in the picture frame. However, not all paper documents or books have the format proportions of the camera photos, so you wil have a background in the view whatever is your zoom, unless you zoom in to see only a part of the original image. With thick books it is advisable to adjust the cradle position every 20 pages or so in order to get the picture without the opposite page getting in the view (the middle part of the book moves a little with the turning of pages).

The background itself is a v-shaped black matt piece of paperboard that I put in the cradle. It is black because some programs make it easy to remove the black background automatically for mono images (it is much more tricky for color images).

For single sheets of documents that are not in color, there is no need to worry about zooming and background. The black border background can be taken care automatically in batch in the processing stage.

As a matter of fact, I am too impatient at the photo taking stage to adjust the cradle every so often. I let the black background margins and a fragment of the opposite page to be in the frame and I cut the margins in the processing stage. For that (cutting margins) it is important that the book pages are always in the same place in the cradle. Therefore, I use a "delimiter", a v-shaped stiff black piece of carboard one or two centimeters wide, one or two millimeters thick (v wings about 2 centimetrs high) that I put at the edge of my black backgroud paper in the cradle, and I always push the book against it.

Once the position of the cradle, camera zoom and focus are set, I shoot all the odd pages first by turning the pages and flattening them with glass, and then I reverse the book and shoot all the even pages. After downloading the images from the camera, I put the first half of images (odd pages) in a separate folder and the second part (even pages) in another folder. Now I can cut the margins in batch automatically - separately for odd and even images (there are programs that can set the cutting parameters different for odd and even pages in one batch). Next, I rename the images in the folders automatically using Irfanview, so that odd images are 0001.jpg, 0003. jpg, etc., and even images are 0952.jpg, 0900.jpg, etc. The even images are in the reverse order since the book was reversed in the cradle for the shooting. I do not bother with pages being in horizontal or upside down position. My OCR program takes care of it. However one can rotate all images in the batch (separately for odd and even pages) automatically with Irfanview of Picasa (recommended).

Now, the two batches go into one folder, and can be processed by OCR or any other program and converted to pdf.

For small booklets or magazines I do not shoot odd and even pages separately but I reverse the original after every shot. Than I crop margins manually in Picasa.

For single sheet documents the whole workflow is much more simple. Just taking a shot a replacing the sheet. Five seconds per page. After downloading images from the camera (Picasa recommended) you can review them and improve by cropping and fine tuning if you wish.

BTW. Apart from anything else, the scanners are much too slow for book copying. They are really good for a batch of single sheet documents fed by an automatic feeder.

For that (cutting margins) it is important that the book pages are always in the same place in the cradle. Therefore, I use a "delimiter", a v-shaped stiff black piece of carboard one or two centimeters wide, one or two millimeters thick (v wings about 2 centimetrs high) that I put at the edge of my black backgroud paper in the cradle, and I always push the book against it.

In one of the next post I will show how to take photos with a small compact camera not connected to the computer.

[QUOTE=Studio717;101293]Can Do you use the glass on each page? /QUOTE]

For really fast shooting, I set the focus of the camera and a time of i.e. 10 seconds for automatic shooting. Then, I only turn the pages and keep them flat with my fingers. The fingers in images can be later removed by trimming the pictures in batch. A book per hour is possible.

ereszet,

Thank you for this! I'm considering making a cradle out of foam core. Plenty of strength and yet easy to cut (at least I'm hoping so - on both counts). I have both an old copy stand and a small light tent, so once the cradle is built, it should be a matter of putting it all together and tweaking.

I'm looking forward to your next 'installment'. Thanks for taking the time to do this.

ereszet,

Thank you for this! I'm considering making a cradle out of foam core. Plenty of strength and yet easy to cut (at least I'm hoping so - on both counts). I have both an old copy stand and a small light tent, so once the cradle is built, it should be a matter of putting it all together and tweaking.

I'm looking forward to your next 'installment'. Thanks for taking the time to do this.

A foam cradle should work with two improvements.

1. Glue it to a heavy base, so that it doesn't move easily.
2. Make some wedge panels of different angles (the foam should be ok for that) to put them either under or over the cradle to avoid trapezoidal/perspective effect. The v-cradle should be at an angle greater than 45 degrees to make it possible to use the wedges. The wedges can make the angle smaller but not larger, unless you put them under the cradle at the opposite side of the book page (in that case you can use anything, e.g. for my cheap and dirty cradle I used books of various thickness under one end of the cradle to change the angle.

Attach the pictures of your setup, when it is ready.

For my do-it-yourself v-cradle I am going to add four bolt legs, to regulate precisely the horizontal position of the v-panels to match that of the camera lens. In theory a small level tool should help to get both the camera and the cradle in the right horizontal position, but in practice the camera postion changes a little when you fasten the ball head (unless you buy a precise and expensive 3-dimensional ball head for your tripod or copy stand). That is why a preview on a computer screen is the only way to make sure that the picture is a perfect rectangle.

Can you see the differences in lighting at the margins of the original picture? Hardly so.
But the computer can, as displayed in the decorative frame of the second picture.

For my example how to remove black background see my post #223 in the thread "Yet another PDF to LRF converter" (Reader Content) (http://www.mobileread.com/forums/showthread.php?p=102036#post102036).

I am very interested in your project

can you tell me please some info: what camera you use, what lighting, what equipment, is OCR satisfactory? like scanner's scans?

thanks

If you read from the start of the thread, user, you'll find all your questions answered.

I didnt find anything about the camera

Yes, this thread is kind of a tutorial, but nevertheless it is nice to get a feedback from the readers. So here is summary of some points.

1. Scanner vs. digital camera

Scanner is ok for single page documents, and a scanner with a feeder is very good for a pile of single sheet documents. Scanners can also be used for scanning your old photos to digital form. They do not require any lighting arrangements and take care of the white balance (usually produce white background as white - but not always). However, getting the right colors with a scanner requires as much care and processing as with a digital camera. You can learn a lot about scanning from the author of Scanview and the links on his pages: http://www.hamrick.com/vsm.html

I have a HP scanner on my desk that I don't use anymore (I dont't use a photocopier anymore either). The main reason is that the scanner is very slow in color mode. The mono mode is faster but still too slow for me. The other reason is that you cannot spread flat a book in the scanner even if you use a heavy weight on the scanner's cover. If a book page is not flat, the lines of text will be curved, and OCR will be unreliable.

Sometimes (e.g. for professional printing) you may need a scanner resolution of 1200 dpi or more. Few (very expensive) digital cameras can match that resolution for A4 pages today.

2. Digital cameras

I started with a 1.2 Mpixel digital camera only five years ago. Totally useless. (Even before that I had an opportunity to casually talk about the idea of using digital cameras as scanners with some high level HP executives. They didn't "get the picture".)

My next camera was 5 Mpixel Canon PowerShot G5. Image sensor 1/1.8 inch CCD. Image size up to 2592 x 1944 pixels. Very good.

My current camera is 8 Mpixel CanonPowerShot Pro 1. Image Sensor 2/3 inch CCD. Image size up to 3264 x 2448 pixels. Canon does not want my money for the successor of Pro 1. Apparently it competes with high priced reflex cameras.

Most Canon cameras can be connected to a computer to shoot pictures (not just to download them). Most reviewers negelect this advantage and also do not understand the need for higher resolution and never refer to repro applications in their reviews. While 8 Mpixels are good enough for A4 pages (210x297 millimeters) and more than good (with zoom) for A5 pages (you can calculate the resulting resolution from image size and paper size), it would be good to have as high resolution as possible for A3 and even A2 pages. As a matter of fact with my Canon I can shoot a picture of a distant object over the heads of the crowd and read the text which I cannot see well with my eyes. It is also the only way to read small print instructions that come with some products (recently I used my camera LCD to read a serial number on my Archos 704 wifi). The higher resolution (in terms in Mpixels) cameras should come with larger sensors. Higher resolution with the same size sensor is sometimes a disadvantage for repro applications. A nice feature for repro applicatons would be the manual focus. It is very awkard in my Canon (requires three hands).

When shooting, always use a neutral mode (no color or contrast improvement in the camera - you can always improve it with proper software) and the lowest ISO possible at a given lighting condition. In my experience, it is better to underexpose than to overexpose. Always take care of proper focus. For all practical reasons, you cannot improve on that once the shot is taken.

As for cheaper compact cameras, it looks from the specification that 10 Mpixel Canon A640 or a succesor may be good for repro applications.

There are cameras available with a special text mode. It is just a selection of camera settings. Good for novices.

My other two cameras: 8 Mpixel Casio Ex-Z850 and 10 Mpixel Casio Ex-Z1000 provide the text mode. They have a very nice 9 point focusing screen. Since they use the same size sensor, the 8 Mpixel Casio is better for repro applications. It is more difficult to focus with the higher resolution Casio.

3. Lighting

With the v-cradle design use a bright light as high over the cradle as possible. A diffusor to dissipate lighting and make it more uniform is useful. A soft flash (possibly redirected against the ceiling) can be used for black and white pages. The most uniform lighting you can get is the sunlight outdoors, with the original paper document at a proper angle to the sun..

4. OCR

Abby Finereader 8 is a marvelous piece of software but sometimes it is vicious and puts strange and obscene words in the recognized text (it is just the way it interprets the picture). It is very risky to use the OCR-ed document without correcting the misspellings manually. I never waste time for correction since I produce the output as the picture-true pdf copy of the original with OCR-ed text layer underneath. In that way you can index and search the documents in your computer and always read and print a perfect copy of the original document.

I've done some research on the web about cradles and discovered that rare book libraries use several different kinds, including book pillows. I'm going to do some experiments (I hope this weekend) to see if these alternatives will work for this type of 'scanning'.

My camera is a 7.1 megapixel Olympus C-7070 with a good macro capability. I'm hoping it will work. I have a remote for it, which will help I think, but the software I'm going to have to look into further. No doubt the solution will be PC only, which is always annoying but not a deal killer. (I usually use Macs, but I have PCs.)

For OCRing I usually just use Acrobat. I used to use a version of Iris but there were limitations that I just didn't want to deal with (they are probably gone in later versions, but I haven't looked). I've also used Omnipage in the past, but not at all recently. My biggest issue is that for many books OCRing is a waste of time. The 18th century long 's' totally, uh, messes up the results.

I'll be printing out ereszet's instructions and trying to replicate with on-hand equipment before buying (or building) new stuff. This looks like a great solution to the problem of scanning page by page in a scanner, even one as useful as the Opticbook (also PC only :disappoin ).

A book pillow? Never heard of that. I can imagine a black soft pillow as a "cradle" for books but it would be a daunting task to keep the book pages at the same position for sequential shots. What about a glass to flatten the pages? Can you give more info on that?
As for OCR, you apparently use it for English texts which are so easy that Acrobat Capture (and even free ocr programs) may be ok. I ocr all kind of languages (even old Russian cyrilic) and books as old as 17th century. I tried Omnipage, Iris, Textbridge, an IBM program whose name I forgot, and any demo/evaluation I could find on the internet. There is no match to Finereader (they even offer a version for Gothic Fraktur and Old European Script but it is too expensive for personal use - apparently they sell it to libraries).
OCR is crucial if you search thousands of books for specific terms and purpose, like genealogical research. You just cannot read thousands of books page by page.

Here's the first instance of a soft book cradle, complete with instructions on how to make it:

Velvet book cradle (http://aic.stanford.edu/sg/bpg/annual/v04/bp04-01.html)

And here's the site that sells the book pillow:

Book support bag / cushion (http://www.conservation-by-design.co.uk/sundries/sundries35.html)

There's also the more traditional library cradle:

Acrylic book cradle (http://www.gaylordmart.com/adblock.asp?abid=8606&sid=82B2B66A1E184839BDC8F939DFE873&search_by=desc&search_for=cradle)

That last one does not have adjustable sides, so that would have to be dealt with. (There is another model that has front 'stops' on it to keep a book from sliding.)

The last cradle (I should say cost-effective cradle) was one made for book binders. A sewing cradle is very close to what's needed, though I think the sides may be a bit too steep and, in the ones I saw, the sides weren't adjustable.

I saw some of the more commercial cradles for digitizing and wish they weren't so expensive! The most common one seems to be flat blocks of varying heights, where one side of the book can be raised and lowered depending on where the book is open to. This would be especially nice for thick books.

If I understand the process clearly, the main concern is to keep the plane of the camera lens even with the plane of the page being imaged, along with keeping the book steady. To keep the efficient flow ereszet has worked out, it's necessary to not have to readjust after each image, so those are the two things I keep in mind when evaluating the usefulness of the cradles and pillows, etc.

I really appreciate the work ereszet has done. It's amazing!

(And you're right, I'm only doing English books - with a few scattered French ones - and nothing earlier than the 18th century. After reading your list, I realize I have it easy! :grin2: )

I am afraid I will have to echo what everyone else replied in this thread: mr ereszet your work in marvelous

I was thinking weeks ago something similar to what you made, but a bit different, here is my idea:

first, there are two ways to digitize afaik: cameras and scanners

the camera problem is that cameras cant create an image of the quality of a scanner (low resolution, much distortion, many artifacts, needs specific lighting conditions, specific camera angles, camera needs megapixels 35MP or more to reach scanners' 300dpi resolution) and we need expensive equipment to achieve results as a scanner easily does, however they are very fast

the scanner problem is that pages cant be flat, which results in distortion at the edges and that it takes too much time to scan each page, however the scans are top quality

my idea is about a scanner that will scan books easily, fast and accurate, without damaging the book

it will scan the book with best quality including the area near the binding
it will scan the book in the half time of a scanner, since it will scan two pages at once

this scanner will look as if you take two flatbed scanners and bind them as a book

here you can see a scheme of it:
http://i17.tinypic.com/4yjykpy.jpg

you can imagine it as an A3 scanner (that scans both pages of an opened book) but it will be foldable

the scanner I propose wont damage the book at all, since the book will be in its normal position as when you read it

you may also look at this picture:
http://www.diy.atiz.com/images/new_site/cladle.jpg
http://www.diy.atiz.com/diy_detail_info.php
http://www.diy.atiz.com/images/new_site/component/pic_self-stabilizing_cradle.gif
http://www.diy.atiz.com/diy_components.php?page=V-shaped,_self-stabilizing_book_cradles

it looks like the scanner I propose, but these glasses dont scan, they are only used to eliminate the shadow and to make the bookpages flat so that a digital camera will take a picture of the pages

opticbook solution is not adequate imo, although I havent used it
it is just a simple scanner with small border:
http://www.plustek.com/product/images/SEE.jpg

microtek scanner (http://www.mustang.com.hk/bottons/Microtek/scanner/s280.htm) has the same 'small border feature' plus its slim, so these scanners maybe the appropriates to create the dual scanner I say

it looks simple, however I dont know if it can be implemented and if there will be any benefit

looking forward to hearing your opinion
thanks

I am proud to see that both the velvet and the acrylic cradle can not match my design in terms of precision positioning and flexibility. You can add a soft padding to my v-cradle panels if you wish. With my new improvement (four regulated bolt legs) my v-cradle will allow to change horizontal and vertical angles to any precision required without touching the tripod. And it allows you to shoot very fast without adjusting the postion for every page. Instead of glass you can use your fingers or a clip or a transparent strap.
If I had a few thousand bucks to spend, I would buy an Atiz two camera v-cradle, but my design does the same (with less automation), with more precision and flexibility.

maybe a plate that will stand on an axis which will allow the plate to rotate it self 180o (and the book on it) would make more easy to shoot the next page

although it will be the first time I will shoot a book, I cant say all the work ereszet is proposing is needed

this simple and cost-effective scheme will be more than enough imo, since it enables you to make any adjust in the shooting angle you may ever want to make, but ofcourse please feel free to make any comments

http://i23.tinypic.com/2yuww3n.jpg

Anything similar to what is in your drawing will work. I tried before and suggested a number of substitutes in my previous posts. When I am in my office, I don't bother to go to a copier. I frequently shoot a paper document placed on a table over a black background, keeping a camera in my hand. As I have already mentioned, I run a paperless office. Fifty or so pages of documents that come to my desk every day can go to my camera and my computer in a matter of 15 minutes. Any distortions I can correct later with batch software processing, but I prefer to do the photos right from the very beginning with the v-cradle in my home studio).

For precise positioning and fast shooting of a book you need more than your design. When trying to aim and zoom your camera from a tripod you will learn how difficult it is to adjust the tripod angles precisely (in three dimensions). I did it once and I don't touch my tripod and camera anymore, I can just move my v-cradle forward and backward, left and right, up and down, change the angle of v-panels (in the next model, even precisely change the length of micrometer legs). Also remember that you need to obscure your camera behind a black curtain or a paperboard, otherwise the camera will be reflected in a glass or a glossy surface of the (color) paper document/book.

A rotating plate was my idea for shooting two opposite pages one after another. A marble rotating plate for serving cheese can be all right for that. However you would need a v-shaped glass and a book with wide margins between the pages, beacuse a joint of the glass would obscure the middle part. What's more, turning the rotating table 180 degrees will take about the same time as turning a page. On top of that, the positioning of opposite pages will be different for books thicker than pamphlets. So I dismissed the idea, and I shoot all the odd pages first and even pages next.

nice

since you are more experienced, can you post here some:

1) program macros for batch image correction (what to correct)
2) calculations for the optimum angles, camera distances, amount of lighting (I suppose prof photographes may help)

something else, your camera costs over 1k USD, will cheaper 12MP cameras will do the job? anyone tested?

also, has anyone tried to work with any of these (http://www.amazon.com/s/ref=nb_ss_p/103-1905967-0107833?initialSearch=1&url=search-alias%3Dphoto&field-keywords=copy+stand&Go.x=0&Go.y=0)?
they are supposed to work well, since many industrial strenght book scanners, look like these, eg (http://www.imageware.de/)

last, I suppose all people occupied with book shooting have tried the snapter program from atiz, JFYI

the camera problem is that cameras cant create an image of the quality of a scanner (low resolution, much distortion, many artifacts, needs specific lighting conditions, specific camera angles, camera needs megapixels 35MP or more to reach scanners' 300dpi resolution) and we need expensive equipment to achieve results as a scanner easily does, however they are very fast
.....
this scanner will look as if you take two flatbed scanners and bind them as a book

With 10 Mpixels camera you get over 300 dpi for A4 format, over 600 for A5, and over 1200 dpi for your collection of small print business cards.

Cost is minimal. Just a tripod and do-it-yourself v-cradle. Today, you need a digital camera anyway on all kind of occasions.

A v-shaped scanner plus automated turning of pages (mechanical or pneumatic) would be a breakthrough. HP already offers some see-through scanners, and that technology would be appropriate for the v-shaped design. I hope somebody from HP or their competitors will read this and make it work (you should better go to the Patent Office before them).

With all that, the v-cradle will be a choice for home and non-corporate users, and will help you copy even large size documents or newspaper pages without spending a lot of money.

nice

since you are more experienced, can you post here some:

1) program macros for batch image correction (what to correct)
2) calculations for the optimum angles, camera distances, amount of lighting (I suppose prof photographes may help)

something else, your camera costs over 1k USD, will cheaper 12MP cameras will do the job? anyone tested?

also, has anyone tried to work with any of these (http://www.amazon.com/s/ref=nb_ss_p/103-1905967-0107833?initialSearch=1&url=search-alias%3Dphoto&field-keywords=copy+stand&Go.x=0&Go.y=0)?

last, I suppose all people occupied with book shooting have tried the snapter program from atiz, JFYI

For batch processing, I am posting some hints and problems in the pdflrf by cacapee thread in Content rather Accessories forum.

You do not need any calculations for angles, lighting, etc. You just shoot your photos and compare results.

My Canon Powershot 1 is today less than 500 USD (I guess), Canon A640 is even less. My two compact Casios are good enough as well, alas there is no way to connect them to a computer for shooting and they have no remote.

I have Bogen/Manfrotto copy stand at home away from where I am now, and I am going to use it instead of tripod with my v-cradle when I am back home.

Atiz hardware looks very good, and Snapter is a promising tool but it is too slow and buggy in version 2.0

Picture 1. A flexible cheap arm bought in a photo accessories shop, attached to a chair. I do not recommend that particular type of arm - it is neither flexible or stiff enough even for small cameras.

Picture 2. Details of a wired remote (again a computer accessories shop) attached to the camera.

Picture 3. Professional flexible arm by Manfrotto (it is for lighting support rather than a camera). It is very solid but costs a lot.

I tried all that some years ago, and I recommend buying a good tripod whith short legs (if you want to keep it on a table together with the cradle) instead.

I almost forgot about them, they are really nice. I use them occasionally.

1. A Manfrotto super clamp with a ball head. You can attach it to almost anything: a shelf, a table, a chair, a lamp stand, etc., and you can take it with you anywhere. It is very solid and stable. I recommend a Manfrotto Photo Catalogue available on the internet. You will find there all kind of photo supports and accessories, including precise ball heads.

2. Same as above plus a quick removal plate

3. A microphone stand (image from the internet). You replace a microphone holder with a ball head and get the best camera "monopod". It does not take your precious space with three legs like a tripod. You can put it on the floor. With its long arm you can reach any place you wish. You can even put your camera very high over the floor with the arm set almost vertical to shoot large posters.

Mic stand! Great idea. Just hit up my musician dh for a boom mic stand. Now I have to find the camera tri-pod mount adapter and I'm in business. :D

Mic stand! Great idea. Just hit up my musician dh for a boom mic stand. Now I have to find the camera tri-pod mount adapter and I'm in business. :D

As far as I can remember (I have no mic stand at hand now), you will need one of those adapters to change the screw thread and size from 1/x" to 1/xx", depending on your mic stand and ball head for the camera.

I run a paperless office.

I don't know what kind of office you run, but I doubt one will ever manage to operate an office paperlessly (well, in the near future at least)

things like seals and signatures in original papers will hardly ever be digitally substituted

I don't know what kind of office you run, but I doubt we will ever manage to operate an office paperlessly (well,, in the near future at least)

things like seals and signatures in original papers will hardly ever be digitally substituted

We will have true paperless offices in the future but for now "a paperless office" is something as described in wiki: http://en.wikipedia.org/wiki/Paperless_office

For me, a paperless office is the way of processing information in the office. I store all the original stamped and signed documents somewhere in the office (it is the job of my secretary) but I never have to search for them physically if I must refer to them. I just use a few words that describe the subject and I retrieve in my computer a list of indexed files that contain those words (a boolean search is possible). I don't even assign any keywords to my document files, which is a standard practice in corporate "paperless office". Every OCR recognized word in my documents is indexed, so a combination of words and date of the file let me find anything in seconds.

To give you an idea of the volume of information I process in my paperless office. Just one of the folders in my computer contains: over 100 thousand documents (over 2 million word count) taking about 7 Gb of disk space plus 300 Mb for the index files. My 500 Gb disk is already nearly full with all kind of info. Of course I use another disk as a backup. This year, one of my disks failed, but I only lost about 10 thousand pictures from my trips abroad (I didn't bother to backup them in time - now I repent that).

To be true, not all documents come from photoscanning. A lot of them are files recerived by e-mail in digital form or conference papers, government regulations, etc. downloaded from the web.

Big size:
- A laptop with a "brick" power supply. A Linksys wireless and a 3G telephone card. A tablet laptop would be nice but for the time being the prices are disouraging.
- Canon camera that can upload pictures via USB directly to the laptop. Another "brick" power supply for Canon.

Small size:
- Sony Portable Reader for crucial b&w documents converted to lrf or png (and some books for leisure). 2 Gb memory card. No power supply necessary for browsing 7000 pages.
- Archos 704 wifi for crucial color documents converted to pdf or png (and some movies for leisure). 80 Gb hdd. A small power supply (battery life is for about 5 hours).
- A Casio compact camera with a 2 GB SD card. No power supply needed for a few hundred shots - enough for a week long business trip.
- A small external SD card reader to move the pictures from the camera SD card to Archos (for Sony the SD card itself with all the pictures can be inserted directly).

In addtion, for both big and small offices on the go one needs a cellphone and a small power supply again.

A bit of advice:
When cosidering a compact camera for the small office-on-the-go try to find a camera with direct usb connection, unlike Casios that are very compact but require a cradle and a power supply.

I hope that in a year or two a color e-ink tablet with 7" screen, long battery life and 3G (etc.) connection will replace dozens of today's devices using "crippled operating" systems..

can you tell me please what the last 2 photos here show?
http://www.mobileread.com/forums/showpost.php?p=98093&postcount=3

how did you achieve the last photo here? it looks like scanned, is it an actual photo? its very OCRable
http://www.mobileread.com/forums/showpost.php?p=99347&postcount=13

last, how did you shoot the second of these photos? its very OCRable
http://www.mobileread.com/forums/showpost.php?p=102042&postcount=28

thanks

can you tell me please what the last 2 photos here show?
http://www.mobileread.com/forums/showpost.php?p=98093&postcount=3


It is a foldable light diffusor (photo accessories shop). When folded it is ca. 30cm in diameter. When unfolded it is over 1 meter in diameter. It is attached with strings and clamps to a lamp post.

how did you achieve the last photo here? it looks like scanned, is it an actual photo? its very OCRable
http://www.mobileread.com/forums/showpost.php?p=99347&postcount=13
thanks

Yes, it is an actual photo of a test page, to show that nice final results are achievable. I cannot remember now whether I took the DayTungsten.jpg and cut the margins or did I took another shot with a zoom to get that part of the page in the frame without black borders. I may have used Picasa to automatically improve contrast and perhaps add some brightness.

All that has little to do with OCR. It is rather for my esthetic pleasure. Finereader recommends not too manipulate the images before OCR processing.

In my experience, the only case when some preprocessing helps to get better results in OCR is when the original images are so dark that it is difficult to see details. Picasa's automatic contrast does the job.

In Finereader processing I recommend to change the resolution of the picture to what it really is and not what some cameras provide wrongly in the picture info (like 96 dpi). I usually change the resolution of all pictures in batch to 280 dpi (Canon Powershot 1) or I look at the picture in the Finereader full screen Window and change the resolution in such a way as to get the view of the picture as wide as the window is. I cannot explain the theory behind the last approach but it looks that if I can see the picture well than the OCR processing is more happy.

last, how did you shoot the second of these photos? its very OCRable
http://www.mobileread.com/forums/showpost.php?p=102042&postcount=28


Automatic black border removal and conversion to mono by ClearImage demo.

I convert color pictures of b&w pages to mono in order to reduce the size a picture takes on the disk. With thousands of document pictures stored on a disk, the size matters.

Finereader OCR can automatically convert color to mono on input (it is a legacy feature) but it prefers to deal with original color pictures.

Finereader recommends not too manipulate the images before OCR processing.

this is very interesting, but I cant imagine what kind of logic it has, since manipulating image will make OCR easier

perhaps, ABBY recommends this because finereader already does the manipulation by its own? but in this case, I can't see how pre-OCR manipulation can harm

as for these photos:
http://www.mobileread.com/forums/showpost.php?p=102042&postcount=28
what improved the second/right image and made it perfectly OCR-able?

this is very interesting, but I cant imagine what kind of logic it has, since manipulating image will make OCR easier

perhaps, ABBY recommends this because finereader already does the manipulation by its own? but in this case, I can't see how pre-OCR manipulation can harm

as for these photos:
http://www.mobileread.com/forums/showpost.php?p=102042&postcount=28
what improved the second/right image and made it perfectly OCR-able?

Yes, finereader does in batch: rotating, correcting resolution, straightening text lines (poor algorithm), despeckling, and of course recognition of blocks and text. On input it allows to detect image orientation, split dual pages and convert to black and white.

As for preprocessing they especially advice not to fatten or otherwise manipulate the pixels around the text.

The last question I already answered in the previous post.

JFYI, a nice OCR article aiming to specify optimum OCR conditions (http://groundstate.ca/ocr)

JFYI, a nice OCR article aiming to specify optimum OCR conditions (http://groundstate.ca/ocr)


The approach to testing Linux OCR programs in the link you provided is rather simplistic. It does not take into account that the scanned pages are frequently not "clean" and come with lower resolutions (like fax pages), they contain layouts with blocks of text and images, the text may be in different colors (the author didn't even test italic and bold print for his few selected fonts), the 8 point and even 6 point fonts are not uncommon in scanned originals, the original lines of text may be skewed and even curved, etc. One accented word he used to test the OCR performance (presumably with the French and English languages selected) does not give a clue about recognizing text with a number of different languages in one document (I frequently recognize by OCR more than six languages in old books - Latin, English, French, German, Cyrillic, various Slavic languages, plus their old varieties). While providing some hints on the performance of a few Linux OCR programs, the tests have little to do with real life situations. I am afraid that Linux OCR programs are about 20 years behind the commercial Windows OCR programs, like Finereader 8. It is not about programming, it is about recognition, cleaning, deskewing, binarization and other algorithms.

To give you an idea how correct is the recognition using photoscanning and Finereader 8, I used the line of text as in the tests above, and added 8 point plus italic fonts. Below are the results.

Number of characters (with spaces) 1852

1. Word document converted by software to jpg at 300 dpi (no scanning), picture attached - 6 uncertain characters (no errors)
2. Word document converted by software to jpg at 400 dpi (no scanning) - 16 uncertain characters
3. Paper printout of the Word document photoscanned at about 300 dpi, picture attached - 17 uncertain characters
4. Paper printout of the Word document photoscanned at about 300 dpi and binarized by ClearImage demo, picture attached - 11 uncertain characters

While most uncertain characters were recognized correctly, some where wrong - mostly wrong recognition of characters in the articles "the"). The 1% uncertain/error margin will make little difference when searching for a combination of words in an OCRred and indexed files, however it would make a difference for presenting the document as a copy of the original. That is why for the output of OCR the best option is to use pdf picture over text.

As strange as it may look, 400 dpi resolution did not improve the recognition but binarization did.

In the photoscanned picture you may notice that corners are a little darker that the rest of the page. This is a feature of camera zoom. To avoid dark corners you just use a lower zoom. The higher is the resolution of camera sensor, the lower is the zoom required to get the photo at a given resolution for a page of a given size.

may film cameras have better results than digital cameras?

may film cameras have better results than digital cameras?

I guess that a film camera (i.e. still photo film camera) can produce a better result (in terms of resolution and color) than a digital camera. Only a few years ago, any professional photographer would dismiss digital camera as a toy. But today's digital cameras are a different breed, so we need an opinion of somebody who still uses a reflex film camera.

As far as examples of microfilm reproductions available in digital libraries are concerned, the quality of pictures is very bad. Is it because they didn't have a repro v-cradle for shooting? :)

If you need only paper copies of paper documents and not digital copies, you may achieve the same (or better quality?) with my v-cradle design and a film camera, but the lighting, focus and postioning problems will be the same for analog and digital shooting, with much more difficult and costly setting and processing for analog pictures.

If you think of a digital video camera, the zooming is impressive (x800 etc.) but when you save a frame, the resolution is rather low. However, you can probably use the zoom to read your neighbour's book over a distance. :)

maybe printing the digital camera shoots and then scan them with a sheetfed scanner will be increase OCR accuracy and document quality?

maybe printing the digital camera shoots and then scan them with a sheetfed scanner will be increase OCR accuracy and document quality?
If you prefer to buy a camera instead of a house, you may use a Hasselblad or similar "space age" film camera, print the photo at 1200 dpi and scan it. Perhaps it can improve the rendering of 6 point and lower fonts in case you are trying to make a copy of a thumbnail size Bible. But you can use a digital Hasselblad H3D for that as well.

As you can read in my previous post, even the perfect copy of a Word document will give you some warnings/errors at the OCR stage. You can correct them manually and you can teach the OCR program to recognize most frequent warnings/errors correctly, if you ocr documents with a similar typeset.

Generally, I do not see a reason for you to concentrate on high-end cameras, scanners or expensive repro setups, unless you are in a business of printing, cartographic, geodetic, CAD or similar industry.

For a paperless office of any size a v-cradle, a proper lighting setup and a "descent" digital camera will do the job. The students can have their texbooks converted to Sony Reader in a matter of hours (copyright observed - it is different for different publications and countries) and after processing with pdflrf they are perfectly readable on a small and greyish Reader screen. The OCR and indexing will help you find any word or phrase if you store the textbooks in your laptop (unfortuntely, the Sony Reader will not allow you to search for text.

as for ABBY finereader, while it does very good ocr job, its spell checking is mediocre:
you can't edit the dictionary entries
you cant use regex to implemet your own correction rules of common mistakes
it doesn't autocorrect 'on the fly'

as for ABBY finereader, while it does very good ocr job, its spell checking is mediocre:
you can't edit the dictionary entries
you cant use regex to implemet your own correction rules of common mistakes
it doesn't autocorrect 'on the fly'


In my Finereader 8 (version 9 is long overdue) you can use a Language Editor to create your own language based on one of the existing languages (i.e. copy of English) and add/import your own "user dictionary". There is also an option to use regular expressions. You can choose characters for your recognition set, prohibited characters, characters ignored in words. There are various options for punctuation characters and a Pattern Editor as well.

Having said all that, I admit that I have no time to play with all the options available. My current backlog of documents to ocr is over 1000 pages (and several hundred downloaded books). The 1% error margin in word recognition is acceptable for indexed text search, when you can use alternative words (or wildcards) to find and display a document with the words in context.

I owe you - the visitors to this thread - a warning. While photocopying and processing images is fun, the fun ends when you have to save and file the final pdf. You must give a meaningful name to every document and file it in a proper directory/folder. It is easy for a 500 page book. It is boring for 500 single sheet documents.

I always save my documents with a name starting with current date in the format 20071012 xxx yyy zzz, where xxx etc. are meanigful keywords like "letter from ..." etc. Although it is easy to find an indexed document by searching the key words (key words rather then keywords - adding keywords would be another boring and time consuming step), sometimes you find thousand of documents referring to a similar subject. The date helps to find the document you are looking for, sometimes even without the index search but by simple looking at or searching the names in a directory. Total Commander is a very useful tool for that.

Unfortunately, we cannot avoid this time consuming manual step of giving names to digital documents unless we delegate this job to our secretaries. I do it myself, since I am the only one who takes decisions on processing the document contents. The secretary classifies and files the documents in a traditional way, but we rarely need to rertrieve them physically from paper document folders.

Fortunately, most of what I'm saving is research materials, so I give the files names I can later search for. The hard part is remembering to add in keywords.

A cradle I ordered is scheduled to arrive Monday, so I'm hoping I can test out my system next week. (The adapter for the boom mic stand arrived today.) Getting close!

any recommendations about the glass used to flaten the pages? what material, coatings, dimensions, density, thickness?


A cradle I ordered is scheduled to arrive Monday, so I'm hoping I can test out my system next week. (The adapter for the boom mic stand arrived today.) Getting close!

may I ask where you order the cradle from?

[QUOTE=user;105954]any recommendations about the glass used to flaten the pages? what material, coatings, dimensions, density, thickness?

My glass is about 5 millimeters thick and heavy enough to flatten even creased magazine pages. The size is A4 (210x297) plus 1 cm margin on each side. It is a kind of glass used for glass shelves, nothing special. No anti-glare glass is required with my setup, but possibly with an anti-glare glass you could avoid obscuring the camera with a black paperboard. I have tried a glass from my damaged scanner but it was too light to flatten the pages. Good enough to flatten my family photos to digitize.

as for my scanner suggestion, I think we are close:

the HP 4670 scanner is only 0.75 inches thick -> scanning of book pages while having the book opened as minimum as possible
the Microtek s280 scanner has almost no margin (the plastic margin outside the glass scanning area) -> full page scanning without any distortion near the book spine

if we could combine the features of the two above, we will have a scanner that can be placed between book pages almost as a paper and it will scan the pages while having the book almost closed and without any distortion near the spine

as for my scanner suggestion, I think we are close:

the HP 4670 scanner is only 0.75 inches thick -> scanning of book pages while having the book opened as minimum as possible
the Microtek s280 scanner has almost no margin (the plastic margin outside the glass scanning area) -> full page scanning without any distortion near the book spine

if we could combine the features of the two above, we will have a scanner that can be placed between book pages almost as a paper and it will scan the pages while having the book almost closed and without any distortion near the spine

One company that is innovative and capable of constructing a v-shaped scanner based repro station is Atiz.

They already offer an automatic book scanner called BookDrive and a manual photo scanner called BookDrive DIY, so they have knowledge and technical skills to combine the two in a v-shaped scanner. But don't hold your breath yet.

Each solution has its advantages and disadvantages. The BookDrive is automatic but is only suitable for a certain size and thickness of books. The BookDrive DIY is semi-automatic (the v-cradle centers automatically, the shots can be taken automatically, but requires manual rasing of transparent cover and manual turning of pages). Again, the size of books is limited to the design dimensions. To make a v-shaped scanner competitive one would have to use the BookDrive DIY design with full automation, which means additional mechanisms and costs. Scanners are rather slow in scanning (especially higher resolutions and color pages), and automatic raising of the v-shaped scanner would take even more time. So the advantage of scanning 700 pages an hour (as Atiz claims) would have gone. The cost would be in the range of several thousand dollars if not more.

For a home and small office users my flexible v-cradle design has all the advantages with one disadvantage - the manual operation. There are no limits as to the size or thickness of books (you can put a big size cardboard base over v-cradle panels) and the shooting is as fast as you are in turning pages and flattening them with glass (the resolution and color doesn't matter for the speed of camera shooting). Once you set your lighting conditions right (you may experiment with a black tent to prevent any extra lights or shadows) the quality of reproductions is good enough even for professional applications. Several hundred OCR quality photos per hour is possible.

One company that is innovative and capable of constructing a v-shaped scanner based repro station is Atiz.


a slim-and-without-margin scanner is what will do the job perfectly in every occasion, no need for V shape, cameras, etc

consider the glass you put on the book to flaten the bookpage, to perform a fast scan <3 seconds and produce a 600 dpi image, wouldnt it be marvelous? its the easiest way

I cant imagine any other more efficient and convenient way to produce high quality copies

the digital camera solution is tempting, but I suppose it needs too much work and the result will never reach even 300dpi (in all digital camera forums, in all digital camera manufactures, I contacted, I was adviced to use scanner instead of a camera)

although, I am about to buy a Canon G9 (12MP - 6x optical) and make some attempts, but I doubt I will ever reach kirtas-tech.com or atiz.com results since they use Canon EOS of at least 17MP

the digital camera solution is tempting, but I suppose it needs too much work and the result will never reach even 300dpi

You are wrong about dpi. Firstly, you do not use 600 dpi for OCR, unless your print is very small. Recommended dpi is 300-400 dpi, and in one of my previous posts I gave you an example how a perfect (Word print to image) picture gives better OCR results at 300 dpi rather than 400 dpi - with Finereader 8. Faxes at 200 dpi are OCRed with no problems as long as they are clean (no smudges).

Secondly, a 10 Mpixel camera gives you an equivalent of 300 dpi for A4 format and 600 dpi for A5. I refer you to a nice OPTICAL DPI TABLE in Atiz FAQ (http://www.atiz.com/support.php). You can easily calculate it yourself as well. However, to be true, I quote below a comment by Atiz:
"It is kind of hard to compare the dpi from cameras vs. the dpi from scanners. The results from the cameras at lower dpi can be higher quality than scanners. DPI is not the total information that is equal to the quality of the image. It is just one factor of many other factors that represent quality."

I agree with them. I have used a dozen of various scanners (including sheet feeder scanners) in my work. One of them is standing idle on my desk even now, but I have no use for it. Photoscanning is fast and produces good quality results for normal home or office use.

mm what about barrel distortion, white balance, color distortions that are common in cameras?

scanners have minimum artifacts (mainly from dust) and inexisting distortion

I dont have an opinion about camera scanning at the moment, I just collect facts and most photographers say it wont work or it will be too much job for such a bad quality

I hope G9 will be enough and I will come back with positive feedback

the atiz supported cameras are +1k USD

mm what about barrel distortion, white balance, color distortions that are common in cameras?

scanners have minimum artifacts (mainly from dust) and inexisting distortion

I dont have an opinion about camera scanning at the moment, I just collect facts and most photographers say it wont work or it will be too much job for such a bad quality

I hope G9 will be enough and I will come back with positive feedback

the atiz supported cameras are +1k USD

1. I see no barrel distortion when shooting A5-A4 formats. I see sometimes darker edges, which can be avoided by changing the camera zoom. White balance is the problem of lighting and your perception. All the digital cameras I have, provide white balance settings for several different light conditions (daylight, fluorescent, tungsten, automatic) and a possibilty to set the parameters based on your shooting of what you believe is a white page (or a Kodak gray sheet). If your light is uniform you can easily correct white balance in RAW photos or even in jpg (see Picasa). Color depends on the mode settings in the camera. I recommend neutral (no processing inside the camera) mode. You can match the colors to your taste in a number of photo processing programs (Picasa included). Scanners need as much color attention and postprocessing as digital cameras. And yes, the scanners (especially the ones in the range of $100) distort images, because of non uniform speed of their CCD or other sensor bar. As for white balance, most of the time the white from a scanner looks white but not always, and you have no means to adjust it in the scanning process (I do not refer to high end costly scanners)

2. Most photographers dismissed digital cameras a few years ago. Times change.

3. Your Powershot G9 should be more than enough for the job. My first digital scanning camera was G5 (5 Mpixels) and was quite good for A4 format sheets and OCR. Unfortunately, Canon makes some strange marketing moves. They removed raw mode from G7 (fortunately it is back in G9), but they also removed a flexible, rotating LCD screen (which is still available on a cheaper A640). That, and also the lack of a pop-up flash in G-series, is the reason why I stay since 2004 with my Powershot Pro 1 and wonder why Canon doesn't want my money for Pro 2.

4. As for Atiz DIY BookDrive, they provide support only for selected cameras, because they are not in the camera business. Canon cameras come with the software to shoot from the computer, and some come with infrared remote. An independent developer of Breeze Systems sells the PSremote program that works with practically all Canon cameras, and is much more convenient than the Canon software.

It is time to present some comparison results scanner vs. camera.
The scanner is hp scanjet 2300c at 300 dpi color.
The camera setup is Canon Powershot Pro 1 on a tripod, v-cradle put on a box to bring it closer to the lens, a home 50 watt lamp with halogen mini bulb - about 1.5 meters over the cradle, a diffusor for dissipation of light - about 40 centimeters below the bulb (see my first posts in this thread).

Neither the scanner nor the photo setup are top quality devices/arrangements. It is just what I have at hand in my temporary residence away from home.

The scanning took a lot of time (between 30 and 60 seconds if not more - I didn't look at my watch).
The photo took a split second plus a dozen or so seconds to download the image to my computer. All camera parameters apart from zoom and white balance were automatic. White balance was set to "tungsten". No attempt was made to set the custom white balance, based on a Kodak gray card (or a white sheet of paper).

The object to scan and photo was a DVD plastic reflective case with a sleeve cover underneath and miniscule text in 10 languages. See the attached picture "original scan and photo.jpg" (lefthand image is scan, righthand image is photo).

The black margins from both the scan and the photo were removed manually. See the attached picture "images with black margines removed.jpg" (lefthand image is scan, righthand image is photo). The colors are not true neither in scan (too blueish) nor in photo (too greyish). The white is more white in the scan (possibly can be made better in the photo with custom white balance settings or with postprocessing). The photo shows a small barrel distortion due to the object being close to the lens. My camera would not allow me to zoom from a further distance (automatic focus would not work; I have not tried to use the manual focus since it is awkard and cumbersome to use in Powershot Pro 1 - despite a nice ultrasonic ring that can be rotated manually, one needs to push some buttons at the same time, which requires three hands).

The quality of images can be seen in detail in the attache picture "scanner and photo detail.jpg" (top image is scan, bottom image is photo). Amazingly, the photo renders the fonts much better than the scan.

The OCR results with Finereader 8 are (out of 10 languages, two were not installed in my recognition engine):
scan - uncertain characters 82, total characters 2130; error rate 4%
photo - uncertain characters 44; total characters 2093; error rate 2%

To make the OCR results visible, I saved the results as text and image. See the attached "text and image picture" (lefthand image is scan, righthand image is photo - both after OCR). Usually, I save text under image, which preserves picture true copy of the original.

What arrangement wins? In my opinion, both for speed and quality the camera is better. But you can draw your own conclusions.

As a bonus, I include an example of OCRed photo of text on a Scan Reader screen (original at the top, OCRed "text over image" at the bottom of the picture). That the result is perfect is just accidental. You can expect some OCR errors with other texts.

thank you very much for this!

the results are astonishing to me, camera way wins hands down, far superior to scanner way

however, I think Canon Powershot Pro 1, as a professional SLR camera, even if the megapixels seem low for todays standards (8MP), it is better than some of todays compact/ultracompact 12MP cameras, do you get the same results with your compact casio?

here (http://www.dpreview.com/learn/?/key=sensor%20sizes) you can see why SLR cameras are superior to compact cameras and why 8MP is not the same for all cameras

so, it would be interesting to estimate the optimum camera that would do the job and that anything more would be an overkill

in "original scan and photo.jpg" the scanned image is at the left? why its so small?

in "original scan and photo.jpg" the scanned image is at the left? why its so small?

The object in the original scan is placed in an A4 scanner so it looks small. The object in the original photo is taken with the maximum zoom allowed by the camera (while retaining focus) to get as large the picture in the frame as possible.

The scanner set at 300 dpi will make 300 dpi scans no matter what is the size of the object, be it an A4 document or a 1" stamp. The camera can zoom to get the object in the frame with little or no background, so the dpi equivalent is greater for smaller object (same Mpixels for smaller size). It is a kind of self-adjusting dpi.

One can say that it is unfair to compare a set dpi in a scanner with a self-adjusting dpi equivalent in a camera, which can much exceed the scanner setting. But my comparison is not about fairness but about practicalities. I will not even try to waste my time waiting for the scanner to take a color scan at 600 dpi or higher resolution. It would take ages.

As for my camera, Powershot Pro 1, despite the name, is not a professional SLR. It is an old "prosumer" camera belonging to the same class as Canon G-series cameras (the sensor is much smaller than the one in SLRs). Later today, I will conduct some tests with the same object and my compact cameras.

Later today, I will conduct some tests with the same object and my compact cameras.

this will be very interesting

the sensor is much smaller than the one in SLRs

the sensor size of Canon PowerShot Pro1 is 2/3", which is significantly bigger than most compact cameras' sensor size (1/7", etc)

and given the fact that, "digital compact cameras have substantially smaller sensors offering a similar number of pixels. As a consequence, the pixels are much smaller, which is a key reason for the image quality difference", the results may be disappointing

pixel count does not accurately determines image quality, pixel quality (among other factors) defines image quality
image quality may be better in low megapixels, if the pixel quality is better

also, did you save in RAW or JPG?

Following my earlier post, below are further test results including two models of compact Casio cameras.

Setup: since I do not touch my tripod and Canon Powershot Pro 1 anymore to avoid lengthy angles adjustment afterwords, for testing Casio I used a superclamp with a headball, attached to the back of a chair and my v-cradle put on the seat of the chair over a cardboard box. I took the photos outdoors (picture attached).

The settings:
1. HP scanjet 2300c 300 dpi color
2. Canon Powershot Pro 1, 8 Mpixels, resolution 3264 x 2448, sensor ccd 2/3" (8.80 x 6.60 mm), indoors with halogen 50 watt lamp, white balance set to tungsten, other settings automatic
3. Casio Ex-Z1000, 10.1 Mpixels, resolution 3648x2736, sensor ccd 1/1.8 " (7.18 x 5.32 mm), outdoors in shadow, white balance set to custom based on white sheet of paper, other settings automatic
4. Casio Ex-Z850, 8.1 Mpixels, resolution 3264 x 2448, sensor the same as Ex-Z1000, outdoors in sunlight, automatic white balance, other settings automatic.

The picture attached (margins trimmed.jpg) shows in the top row images from the scanner and the Canon, and in the bottom row images from Ex-Z1000 and Ex-Z850. Apparently my custom white balance setting for Ex-Z1000 was wrong (I used a white sheet of paper instead of Kodak gray card. However a white sheet works for me indoors) .

The details in the attached picture "detail.jpg" show that rendering of fonts with camera photos was in all cases better than in the scan (download and see the picture full screen). The images are in sequence 1 to 4 from top to bottom.

OCR results with Finereader 8 (no preprocessing of pictures)

1. scan - uncertain characters 82, total characters 2130; error rate 4%
2. Canon - uncertain characters 44; total characters 2093; error rate 2%
3. Ex-Z1000 - uncertain characters 52; total characters 2100; error rate 2%
4. Ex-Z850 - uncertain characters 80; total characters 2133; error rate 4%

The pdfs saved as text and images are attached in the last picture.

Conclusions:
- The results from a compact Ex-Z850 camera are comparable to those from the 300 dpi scan.
- Canon (8 Mpixels) gives better results than Ex-1000 (10.1 Mpixels) because of a larger sensor.
- The higher resolution (10.1 Mpixels) Casio gave better results than 8.1 Mpixels Casio, despite the same sensor size (however, it is not as good as 8.1 Mpixels Casio when focusing under dim condtions).

BTW. I have no time to deal with RAW, so I save pictures as the biggest size jpgs (compressed in the camera). Typical sizes are 2 to 4 MB.

Very interesting results, imo they prove that high MP but cheap cameras wont work well. The 8MP Canon outperforms the 10MP Casio, so megapixels is not the only factor we must take into consideration: we need an optimum of both amount and quality of megapixels.

although 98% OCR success is superb, what if you postprocess the images with a software? do you reach a 100%? does postprocessing increase the OCR success in both scans and photos?

I am sure your work will be referenced from all those who try to digitize using a camera, and maybe in the futures scanners will extinct for home/small-office digitization

although 98% OCR success is superb, what if you postprocess the images with a software? do you reach a 100%? does postprocessing increase the OCR success in both scans and photos?

I am sure your work will be referenced from all those who try to digitize using a camera, and maybe in the futures scanners will extinct for home/small-office digitization

1. Please note, that my test was a kind of a "destructive test" (like crushing your car against the wall to see damages). The title sleeve was under a reflective plastic cover, the text was miniscule and in ten languages of which two were not installed in my OCR engine, the pictures (flags) were so small that could easily be taken as a text in color, the background was in color, etc. It is difficult to imagine similar case in practice. That the error rate with my camera setup was just 2% was amazing. One should examine the text after OCR (ten languages) by reading it to get the real taste of what Finereader is capable of doing with moderately good photos (no white balance) of extremely difficult layouts.

On average you can expect a 1% error rate (even with picture perfect photos/scans of text), unless you have hundreds of similar pages and you train the program (dictionary, patterns) with the first dozen or so pages to make sure that the rest is recognized error free.

In my experience, there is always an error here and there. Even if you scan the same document twice under identical conditions, the errors will appear in different places. Apart from sophisticated algorithms, some kind of a huge contextual database and artificial intelligence is the future. With handwritten text there are different recognition algorithms based on the movement of your pen rather then patterns.

2. Preprocessing is not required if you take the photos in proper lighting condtions and make sure that they are rectangular (removing of black borders may be required when the book size proportions are different from the photo frame). Batch contrast improvement may be useful for photos taken in dim lighting (in a hotel on your business trip). Deskewing and despeckling may be useful for faxes that you receive (a lot of people do not clean their fax machines). Correcting perspective and straightening text lines may be required for photos taken by hand or photos of double pages of thick books. Binarization will help you reduce the size of resultant pdfs. All this can be done before using Finereader and some preprocessing can be done by Finereader (but Finereader uses default parameters that may not be ideal with your specific documents/books). Finally, sometimes you may need to remove black blobs that are due to non uniform lighting conditions. When you binarize color or grey pictures with some areas where text and background are barely discernible, you will get the text surrounded by black lines, smudges and blobs. By changing the binarization parameters you can get less of that but the danger is that the text will disappear as well. The best method I found is to recognize blocks of text and pictures in Finereader and save all the images with blocks of text and picture only. Then, you load the images back to the Finerader and check the thumbnails. You will easily see those images that still have black spots and you can remove them with the eraser. After that you save again your images as blocks, load them back, recognize the text, save the final result in whatever format you want, and add white margins afterwords if you need them.

Please do not be put off by the procedures I described above. If you use my setup with adjustable v-cradle you will shoot 10 perfect photos a minute (a book in one hour - with automatic camera shooting, 6 seconds lapse time) and you can directly OCR it. If some pages go wrong (e.g. you turned two pages together instead of one, you moved the book a little and the binding shows in some pictures) you just take again the photos of those pages only.

this is a sample from the Casio EX-Z1000 (10MP)
http://69.93.231.164/PRODS/Z1K/FULLRES/Z1KhMULTI3648F.JPG

this is a sample from the Fujifilm Finepix F50fd (12MP)
http://69.93.231.164/PRODS/F50FD/FULLRES/F50FDhMULTI4000F.jpg

in my opinion, the Casio Z1000 is slightly better (even though its 10MP and Fuji 12MP), the letters are more sharp and clear, with less noise (I dont examine color accuracy)

I cant imagine how can Casio (that is not a big brand in digital cameras) produce so fine quality and how can Casio produce so poor performance in your test (its lower than the 8MP Canon, while Casio is 10MP)

I afraid we will never reach with a compact camera, the OCR results we get from a SLR

also, is there any chance to acchieve more than 98% OCR success? 100% maybe? what is needed? more decent camera? more megapixels? more pixel quality? (98% can be alot of if we are talking about many books)

in this nice site (http://www.imaging-resource.com/IMCOMP/COMPS01.HTM) I couldnt find any 10-12MP compact camera that produces better image quality than Casio EX-Z1000 and Fujifilm F50fd (I didnt look for SLR)

I was thinking to buy one between Casio EX-1080 (10MP), Casio EX-Z1200 (12MP), Fujifilm Finepix F50fd (12MP)

but I cant find the "golden mean", the optimum combination of megapixels and image quality

it is known that cameras with more megapixels produce lower image quality photos, the more megapixels the less image quality (the main is noise and sharpness)

so would a 10MP camera perform better than a 12MP camera in OCR?

oh, this one is very sharp too (I think the sharpest):

Sony Cyber-shot DSC-N2
http://69.93.231.164/PRODS/N2/FULLRES/N2hMULTI3648F.JPG

(unfortunately I cant find Powershot Pro1 images to compare)

Do not take my ad hoc tests as a scientific proof. The results may vary from shot to shot. My tests only show that a digital camera can be as good (and much faster) for scanning as a cheap scanner.

For professional comparison of practical resolutions see http://www.robgalbraith.com. The author provides tests for all the cameras reviewed. (The link didn't work for me today because of Gateway Timeout. Probably a temporary problem).

When deciding on which camera to buy remember about a remote capture feature, which is available for most Canon and very few other cameras. Check http://www.breezesys.com/ to see what cameras the author's Psremote program supports. Also, make sure that the camera can be powered by an external AC power supply.

From http://www.steves-digicams.com, I quote:

"RemoteCapture allows you to operate the camera while tethered to a computer via the USB port and includes a live viewfinder option. The Field angle/flash screen gives you access to settings for: Focusing Point, Flash, Macro Focus, AF-assist lamp and AF operation.

The Image Size and Quality is selected at the top. The shutter can be tripped by either a mouse click or a hot-key on the keyboard.

Captured images can be saved directly to the computer or they can be stored on the camera's CF card and then transferred to the hard drive

The Remote Capture program is also capable of interval and timer recording as well as instantaneous. Shooting interval: approx. 1 – 60 min. (1-min. increments) Number of shots: 2 – 100 shots (Maximum number of shots varies according to memory card capacity"

Avoid compact cameras that need a cradle to download images or to recharge batteries (unfortunately that is the feature of my Casios - two different cradles for two similar cameras). Camera+cradle+power supply is no longer compact.

If I were to buy a compact camera today, I would choose Canon Powershot A640 or a successor.

mmm, I dont think Canon Powershot A640 is the best for OCR

here is a comparison of Canon Powershot A640 with Sony N2 (canon at right, sony at left):
http://shup.com/Shup/8128/untitled1.png

in my opinion, canon looks blury and less sharp than Sony's, but with good color reproduction (black is greyish, but it doesnt have purple in it as Sony's and white seems better that Sony's), however I dont know which is better for OCRing and which is more easily corrected with postprocessing, accurate colors or sharpness

also, there is something else that must be taken into consideration:

when we are saying that Finereader needs at least a 300dpi scan, or at least a 8MP photo, I think we mean that the fonts must have a minimun size so that the OCR program will recognize them and it will be able to see the differences between the letters

so, a small font text will need more dpi/MP than a large font text

also, we must calculate as image dinemsions, not the dimensions of the paper (A4 or more), but the actual dimensions of the text column (and maybe a little bit bigger just to avoid the blurry corners than most cameras have)

so since 10MP to 12MP is a 10% increase in the font size, and at the same time a significant loss of detail and more noise, I wonder which is the golden section of MP-size sensor-font size

in my opinion, canon looks blury and less sharp than Sony's

Good research. But I am afraid that with Sony you will miss this - photos attached.

mm, RemoteCapture would be so important reason to buy a camera?
1) to be able to adjust settings from the pc and view the files to the pc?
2) to save images directly to the pc?
(or there is any other advantage I missed)

arent any alternative programs/firmware that can do this for other cameras? or it need hardware support?

cant one just use a 4GB card and transfer the files? (I suppose it will store the whole book)
cant one change the settings from the tripod and view the images in the lcd?

I recognize that all these would be handy ofcourse, but would it justify a compromise in image quality?

mm, RemoteCapture would be so important reason to buy a camera? ...I recognize that all these would be handy ofcourse, but would it justify a compromise in image quality?

Sorry for the poor quality photo of the laptop screen in my previous post. I forgot about PrintScreen. So here you are (pictures attached).

Previewing a photo on the laptop screen and changing zoom and other settings "on the fly" to get the photo right is crucial for me. Resolution is important but lighting and focus "make or break" the photo. You cannot see the quality of the picture in the camera viewer.

If you are cost conscious you have to suffer compromises.

It is time for examples of quality photo scans. For that, I took the photo of a page from an Ukrainian album on Kiev (Kyiv). Please note, that this is only to illustrate some procedures and therefore for copyright reasons I spoiled the photo with an overlayed text. The album is larger in size than a home scanner can take, so the photo scanning is the only way to copy the page. The photo scanning setup was not ideal due to a temporary arrangement that I use now. In particular, the lighting was a 50 watt overhead halogen bulb, ceiling lamps switched on, and a lighting diffusor. An ideal setup would require a bright (possibly special photo reprographic) overhead lamp.

In the first attached picture (image processing.jpg) you can see the original photo, the same photo binarized to black and white, and then the same binarized photo with the color picture inserted. To achieve that result you just copy the picture from the color original and paste it into the binarized one. You can do it with the free Irfanview program.

Why should one use color pictures in black and white pages instead of dealing with original color pages? For two reasons. First of all, a book of several hundred pages scanned in color will produce a set of images/pdf file of enormous size. Secondly, the black print in color pages is not quite black and the white background is not quite white. For comparison see the attached pictures (aleksander ocr pdf color.jpg and aleksander ocr pdf bw and color.jpg). The three images of OCRed photos in both pictures are sequentially the following pdf formats: text under picture, text over picture, and text and picture only.

Finally, I attach printscreens (lrf color vs black and white with color picture.jpg) from Sony Reader's Connect software that show lrf pages respectively for color and binarized (with color pictures inserted) pdf pages (text under image) refined with pdflrf program. Although the large size album with small print is not something you would carry with you in the Sony Reader, this example shows vividly the diiference between rendering color and mono pages in the Reader's lrf pages.

Long overdue Finereader 9 has been released. I will test it and report the results in this thread as soon asr I buy it. In the meantime I quote below some hints regarding photo scanning from:
http://abbyy.com/DLCenter/downloadcentermanager.aspx?file=/fr90/guides/Guide_English.pdf


1. Make sure that the page fits entirely within the frame.
2. Make sure that lighting is evenly distributed across the page and that there are no dark areas or shadows.
3. Straighten out the page if required and position the camera parallel to the plane of the document so that the lens looks to the center of the text being photographed.

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

Shooting Modes

Lighting
Make sure there is enough light (preferably daylight). In artificial lighting, use two light sources positioned to avoid shadows. (ereszet remark: this is not necessary with the v-cradle design; one bright overhead light source and a light diffusor is recommended).

Positioning the Camera
If possible, use a tripod. Position the lens parallel to the plane of the document and point it toward the center of the text.
At full optical zoom, the distance between the camera and the document must be sufficient to fit the entire document into the frame. Usually this distance will be 50–60 cm. (ereszet remark: tripod with a ball head is not preccise enough to set the camera angle and distance to match different size paper documents every time they change; repro copy stand with micrometer plates will save you a lot of positioning effort; with a tripod, once you position the camera, do not touch it anymore, do all the distance and angle positioning with the v-cradle instead).

Flash
Whenever possible, turn off the flash to avoid glare and sharp shadows on the page. In poor lighting conditions, try using the flash from a distance of about 50 cm, or, preferably, use additional lighting.
Important! Using the flash when photographing documents printed on glossy paper causes the worst glare. (ereszet remark: see my post on redirecting the flash to the ceiling; glare does not interfere with the shots taken with the v-cradle setup).

White Balance
If your camera allows, use a white sheet of paper to set white balance. Otherwise, select the white balance mode which best suits the current lighting conditions. (ereszet remark: white sheet of paper may not work in bright sun; see an example in one of my previous posts; try Kodak grey card instead).

Auto focus may not work properly in poor lighting or when photographing at a close distance. In poor lighting conditions, try using an additional light source. When photographing a document up close, try using the Macro (or Close–Up) mode. Otherwise, if possible, focus the camera manually (ereszet remark: set the zoom and preview the focus with PSRemote remote capture program; if a page is almost blank with a few lines of text use another page of text to set the focus and fix the focus.)
If only a part of the picture is blurred, try reducing the aperture value. Increase the distance between the document and the camera and use maximum zoom. Focus on a point anywhere in between the center and a border of the image (ereszet remark: as long as that point is not a blank part of the page).

Question:
I wonder now if we really need so many megapixels or its useless or overkill. And if not, what is the most importan factor to achieve highest OCR rates.

Scanner:
The above is a scan. I dont know how this scan was produced, what scanner was used, what scanner settings were used, I only know its dimensions (A4), its resolution (850x1103, 100dpi) and that its black-and-white.

http://speakingjazz.com/PageSamples/SpkJz2ndEd-DiatPg19.jpg

Here is the scan that was OCRed:
http://shup.com/Shup/8323/107921064.png
Here is the OCRed text:
http://shup.com/Shup/8485/107921054.png
Finereader results:
Uncertain characters: 18
Total characters: 2087
OCR success rate: 99.14%

The Finereader results were superb in my opinion, considering that, as you can see, most of the uncertain characters were some music characters.

This