Alley in Pisa, Italy

by Lars Kotthoff

submit your photo


Hall of Fame
View past winners from this year

Please participate in Meta
and help us grow.

Take the 2-minute tour ×
Photography Stack Exchange is a question and answer site for professional, enthusiast and amateur photographers. It's 100% free, no registration required.

I am using my S95 to photograph text documents (both paper & microfilm) for OCR. I have already asked about the problems of how to best photograph a computer screen for microfilm, but I've been thinking about the best way to photograph text so that the text comes out clear and crisp. Here are my current thoughts, but I would like to hear what anyone with more photography experience than me has to say:

  1. I am using a smaller # of megapixels (around 6-7MP rather than the S95's maximum, which is 10MP), so that the resulting PDF files don't kill my computer when I actually go to scroll through them

  2. Aperture - lower (e.g. f/6.3), so that if the camera is not exactly parallel to the page, the text will remain in focus

  3. Shutter speed - There's a tradeoff here, because libraries are generally lower-light, so I'd like to lower shutter speed increase exposure, but without a tripod, there is a chance of camera wobble

  4. ISO - I am not really quite sure how this would affect it.

  5. White Balance - Again, not really quite sure.

I've been experimenting with different settings on my camera, but I am just learning about photography and would like to know if anyone has any suggestions for how to best photograph text, from books or from documents, that is easy to read.

share|improve this question
3  
Why not just use the right tool for the job and use a scanner? –  ahockley Feb 10 '11 at 16:58
5  
"Why not just use the right tool for the job and use a scanner?" - Because many of the documents that I work with are not scannable. They are either very fragile, or the library does not allow you to scan, or the library charges exorbitant costs to scan upwards of $0.50/page. –  Jason Feb 10 '11 at 17:46
2  
I would like for it to have been said: I hope that your uses are appropriately respecting copyrights, if and as relevant. :) Presuming so, I hope you find your answers! –  lindes Feb 10 '11 at 18:20
3  
"I hope that your uses are appropriately respecting copyrights" - Yes, my work is with documents outside of copyright restraints (I am a historian, mostly studying late 19th century europe) –  Jason Feb 10 '11 at 18:56
2  
@ahockley: another reason not to scan is that it takes a very long time. With a DSLR I can do about 20 pages per minute on a tabletop, but a scanner that fast uses a document feeder and costs a lot. –  Мסž Feb 11 '11 at 5:12

8 Answers 8

If your camera has a black and white mode, I'd try that to reduce your post processing time. Additionally, many OCR programs I've worked with do much better in B&W.

You'll want get as parallel to the page as you can, because in order to reduce camera shake when hand holding, you're probably going to need your widest aperture in order to maximum your shutter speed. The small sensor on your s95 should increase your DOF enough that I'd open the aperture as wide as you can get away with.

Considering you're just going to be running this through OCR, crank your ISO up to whatever gives you a proper exposure with minimal shake. The OCR is much less likely to complain about the extra noise than blurry letters.

White balance, if you're just talking about text, should be for whatever your "indoor" setting for your camera is - but frankly, don't worry too much about it.

Your megapixels may only matter if your document text is incredibly small and you'll need to clearly resolve that level of detail, but I would guess 6-7 is going to be fine for most things.

share|improve this answer
1  
You probably mean "perpendicular to the page". If you're parallel it'll be hard to read the text. ;-) –  Craig Walker Feb 11 '11 at 0:54
    
@Craig Walker - he used the terms "parallel" to the page and I understood what he meant, so I wasn't trying to confuse him. Not to mention the sensor itself would be parallel, as would the viewfinder. It just depends on what part of the camera your referencing. The term was intentional. –  rfusca Feb 11 '11 at 1:39
    
Use the most megapixels you can, the OCR'd documents are what matters and they'll be relatively tiny and as long as the source is all text, unaffected by the camera resolution anyway. I use Abbyy FineReader and it works better on higher resolution images (21MP instead of 10MP, for example). So shoot at 10MP. –  Мסž Feb 11 '11 at 5:15
    
fair enough. –  Craig Walker Feb 11 '11 at 6:19
    
+1 for B/W mode. RGB for text processing is wasting pixels. –  Greg Feb 12 '11 at 2:39

You seem to have a good grasp of the settings already. Camera shake is going to ruin the readability of text more than anything else, so I would optimize that first. If you can't use a tripod make sure your shutter speed is at least 1 over your focal length (35mm equivalent). Zooming out will make this easier. Use image stabilization if you have it.

Aperture can be set fairly wide, just make sure you're parallel with the page. A wide aperture will help with the shutter speed, again blur due to wide aperture is nothing compared to blur due to camera motion!

ISO, whatever you need to get a good exposure. Noise shouldn't affect readability until it gets really bad.

Shooting a custom white balance based on the paper itself would be preferable, but the OCR software should be able to cope with an off white background.

share|improve this answer
  1. I assume you are converting JPEGs to PDFs. I recommend capturing the maximum, and using a program to downsize to the 6-7 MP with sharpening applied after the downsize and before conversion to PDF.

  2. To clarify, you want a narrower aperture for wider depth of field, which is a higher f-stop number. (e.g. you want f/8 instead of f/2). With text for readability, try to have the camera pointed as perpendicular as possible so you don't need to worry about depth of field. It is a good idea to stop down a bit from wide-open to reduce aberrations such as CA that you will see with black on white text. The closer you are to the paper, the shallower the depth of field will be.

  3. This will probably have the biggest effect on readability if you are handholding. If are handholding, you want to select a shutter speed such that it is faster than 1/(equivalent 35mm focal length), or even faster. This depends on what zoom setting you are using on your P&S. If you use the wide end, you can get away with longer shutter speeds, but may get barrel distortion. If you are photographing text documents and not a computer screen, I recommend trying a flash.

    If you can, prop the document up (taped or something so it is flat) so you can rest the camera on something and get as long of a shutter speed as you want. You can also use a beanbag or small objects to angle the camera as it rests. Also, you can get waterbottle tripod adapters for a P&S.

  4. Aperture, shutter speed, and ISO form an exposure triangle. Read about it here. The higher the ISO, the narrower aperture and shorter shutter speed you can use, but the image will be more noisy. I recommend a low ISO (400 or less) for a P&S with text.

  5. Select the white balance for your lighting situation. If you have florescent, use that. If you are using flash, use flash white balance.

share|improve this answer
    
I don't think f/8 is realistic for shooting handheld indoors (unless using flash, you'll need to ask the library for permission first) plus f/8 is going to cause distraction with a small sensor, the sharpness gain is minimal compared to the required increase in shutter time / ISO. –  Matt Grum Feb 10 '11 at 17:05
    
@Matt f/8 is an example to show that he wants a higher number to increase the depth of field (and reduce other aberrations such as CA, which may be visible). I didn't give him guidelines as I am not sure how much light he has. Distraction? You mean diffraction? –  Eruditass Feb 10 '11 at 17:06
    
Yeah I meant diffraction - I'm writing this on my phone which has an annoying habit of autocorrecting my words even when I spell them correctly. CA wont be a major problem (compared to camera shake), I think you'd be best a stop down from wide open, if light is a big problem... –  Matt Grum Feb 10 '11 at 17:13
    
Don't you love technology? Yeah, I'd say one stop from wide open is a good guideline and agree that camera shake would be the crux. –  Eruditass Feb 10 '11 at 17:19
    
That's a really good suggestion to shoot with max MPs, and then then downsample on the computer. I can imagine that sharpening + extra contrast will lead to much sharper text, easier to read. –  Jason Feb 10 '11 at 17:55

I agree with the suggestions for shooting in the highest resolution available and downsize in in computer. Also, choose the highest quality jpeg you can get from your camera, or even shoot raw if your camera allows it, and convert to a tiff or png. Lossy compression in jpeg can create artifacts around the ink/paper boundary that could confuse your ocr software.

In general, what you want to do is use settings that will get the sharpest image you can get. Things like contrast, white balance etc can be all changed in computer. In fact, increasing contrast using the levels tool or the contrast/brightness tool in photoshop or GIMP could help you a lot.

Also, converting to grayscale format will substantially reduce the size of your files. And if you do that, and increase the contrast, you don't really have to worry about the white balance. You may experiment with the white balance, though, some settings may produce better contrast then others (it depends on your lighting).

If your computer can handle it, do all the adjustments (grayscale, contrast, brightness) before you downsize, and after downsizing, apply some sharpening filter to the image.

share|improve this answer

A few thoughts, which I haven't noticed elsewhere, to add to the mix:

  • If you shoot "zoomed in" (more telephoto), it will cause you to shoot further from the page, which will then give you a more consistent image with respect to size-of-text (words in a corner will be closer to the same size as words in the middle; with wide-angle, you might get distortion with respect to this). I'm not sure what the "ideal" setting is, here, and it's a trade-off against required shutter speeds, potentially, just something to think about.

  • You'll want to "over-expose" versus what your camera probably thinks is right by default (unless it's smart enough to guess that this is paper -- I put that in quotes because you don't really want to over-expose it, just over-expose based on metering off of a white page), because the primary tone (the paper itself) will be the main thing the camera tries to expose for, and if you expose the paper as middle grey, you'll have a muddier starting point. If you "overexpose" by a stop and a half to 2 stops (just don't go to the point of clipping highlights), the text should still be nice and dark, but the paper will be nice and bright, making contrast separations (and thus OCR) easier.

  • try to have even lighting -- if one corner of the page is a lot lighter or darker than another, this will make it harder to separate things out into the pure black and pure white that you most likely want for OCR purposes.

  • a "copy stand" is ideal for projects such as this, though I presume beyond what you could easily have in the library... However, depending on the way your library is configured, it's possible there might be somewhere (perhaps a shelf over the desk, or a rod that holds lighting or similar?) that you could place a super clamp with a ball head (here's one, but any super clamp with an appropriately sized (depending on the head) "stud" and some sort of tripod head would do), and your camera then gets to be mounted over the desk (or whatever) surface, so that you can have consistent positioning, and stability in case you need longer exposures. Basically, this would amount to a "poor man's" copy stand.

    You may also want to use a short self-timer or remote (I imagine your camera has a 2-second self-timer mode, and perhaps a custom one which could be even shorter), to avoid camera shake from pressing the button.

That's all I'm thinking of for the moment. Hopefully that's helpful.

share|improve this answer

I have the same job as you landman photo and have done it for years and years. Most of the people don’t know what they are talking about, trust me!!!

  1. The camera on M or manual

  2. Iso 100-200 the higher the# the brighter but it get more grain or noise in the pic

  3. Shutter speed 1/100 to 1/160 (usually 1/125 is best, the lower the brighter but u get more buries or retakes)

  4. Aperture or f2.7 (why would you need a deep field of vision it’s a flat page)

  5. Shoot in color, 6 to 10 megapixels, look through your white balance settings, usually fluorescent 1 or 2, turn off most of the bells and whistle if Ur not sure turn it off

  6. expose pics on comp using acdc

share|improve this answer

Some small suggestions, but I think your setup looks good

  1. Use a tripod. This will remove any problems with camera shake, and allow you to use as long an exposure as you need.

  2. Avoid light falling directly onto the paper at an angle that can reflect straight into the lens and cause a highlight where you don't want one. This usually means making sure the light is coming from the sides.

  3. ISO: As a rule of thumb, you should err towards a lower value, but I am sure that any material difference will quickly become obvious.

Apart from that, I can't think of any ways to improve on your setup. To be honest, I think you've got it.

share|improve this answer
    
Can you suggest a good (and relatively inexpensive) tripod which would be useful, and allow me to flip the camera over so it's facing parallel to the pages I am photographing? I have looked at the Joby tripods and they look promising, but I've heard that they often shift when you hit the shutter. –  Jason Feb 10 '11 at 17:56
    
@Jason, fair point. I have a Joby, which I like, but to stop it moving when you hit the shutter, just use a remote shutter release or use the self-timer. This is actually a good tip for any time you're using a slow shutter speed. The bad news is that any tripod heavy enough to be very stable is not going to be cheap. Sorry about that! (If you find out differently, let me know!) ;) –  AJ Finch Feb 11 '11 at 10:24

It sounds like you've got most of the answers already :-P I'll just add a few comments on your points:

  1. Resolution: you're right, you won't need a lot of resolution here, especially if you're just going to be viewing the images on-screen. I'd probably cap it at twice the max typical screen resolution myself.

  2. Aperture: you're correct with selecting a smaller aperture to get a bit of depth-of-field wiggle room. However, you're probably not going to need a whole lot. Play with the depth of field calculator for your focal length, figure out a reasonable DOF, and then go with the largest aperture you can. (However, make sure you're a stop or two under your max aperture for best sharpness.) You want the largest aperture so that you can have more room to play with shutter and ISO; see below.

  3. Shutter: obviously you want this as fast as possible to reduce shake, while still getting enough light. I've heard various sources claim "shutter speed equal to focal length" to avoid shake; some claim double. So, if you're using a 50mm lens, try to keep the shutter speed between 1/50 and 1/100. Once you have that, try for the slowest speed to maximize the light.

  4. ISO: you should be able to boost this fairly high. The noise might reduce the look of the picture but probably won't degrade its readability too much. What's "high ISO" and "too noisy" depends on your camera. Experiment. This is probably the best way for you to get your correct exposure in low-light, compared to aperture and shutter.

  5. White Balance: for you, it shouldn't matter too much. If your photographed paper comes out looking off-white, does it matter? (Different story if you're photographing books with full-color pictures of course). If you want to correct it in post, you'll have a relatively easy time, since you have lots of neutral-colored paper to work with. I'd probably leave your camera on Auto White Balance and forget about it.

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.