Photographing Documents With a Digital Camera for OCR

Photographing documents with a digital camera works well for large jobs when the camera is fixed on a stand pointing straight down, the light is even from both sides, and exposure, focus and white balance are locked. Shoot high-quality JPEG, slate every folder so the frames stay in order, and check the institution's camera rules first.

When is a digital camera better than a phone for documents?

A digital camera is better than a phone when you have hundreds of pages to capture in one sitting, when the originals cannot go through a scanner, or when every frame must look the same. Its advantages are a fixed mount, locked settings, a remote trigger and batteries and cards you can swap without stopping.

A phone adjusts itself for every picture. It refocuses, re-meters and rebalances color each time, which is helpful for one page and a nuisance for five hundred, because pages from the same ledger come out in different shades. A camera in manual mode takes the five-hundredth frame exactly as it took the first.

The camera has costs too. Its files have to be moved to a computer before anyone can read them, and cropping to the page happens later on that computer. For the choice between a phone and an office machine, see the comparison of a phone scan and a scanner. If a phone is all you are allowed or all you have, the section on photographing multi-page documents at a courthouse or records office covers that route.

At a courthouse, a pleading you cannot remove from the volume is a good job for the stand. Lock the settings, shoot each photograph the same way, and slate the folder so page order survives the card. Those even frames are easier to read and easier to keep as documentary evidence, with authentication helped by a log of the chain of custody for the memory card.

A copy stand, two lamps and a remote shutter

A copy stand is a baseboard with an upright column that holds the camera pointing straight down at the page. Add one lamp on each side, both aimed at the board from about 45 degrees, and a remote shutter release so that pressing the button never moves the camera. That is the whole rig.

Each part solves one problem:

• The column keeps the camera at one height, so every page is the same size in the frame and the lens stays square to the paper. A page shot square has parallel edges and letters of even size from top to bottom.
• Two lamps at matching angles light the page evenly and send their reflections away from the lens. A single lamp leaves one side darker. A lamp near the camera bounces straight back as glare.
• The remote release, or the camera's self-timer, removes the small shake from a finger on the shutter button.

Setting up, in order

1. Put the stand on a table that does not wobble. Level the camera so its back is parallel to the baseboard.
2. Lay a sheet the size of your largest page on the board. Raise or lower the camera until that sheet fills the frame with a narrow margin all round.
3. Mark the corners of the sheet on the board with removable tape, where the institution's rules allow it, so every page goes in the same spot.
4. Switch off the room light nearest the stand if it throws a shadow of the column, and position the two lamps.
5. Turn the camera's flash off.
6. Photograph a page of printed text. Zoom in on the camera's screen at all four corners and the middle. Every corner should be as sharp and as bright as the middle.

If one corner is soft, the camera is not parallel to the board. If one side is dim, move that lamp closer or the other one back.

Without a stand

A tripod with a column that reverses or tilts can hold a camera over a table, but its legs cast shadows and get in the way of turning pages. Your reading room may forbid tripods or supply stands of its own, so read its rules before you pack one.

Bound volumes add a problem the rig does not solve: the page curves into the spine. Support the book so that the page you are shooting lies as flat as it safely can, and photograph one page per frame. What page curl does to recognition, and what can be done afterwards, is covered under curved pages from bound volumes.

Which camera settings suit pages of text?

Pages of text suit manual exposure with a middle aperture, the camera's lowest standard ISO, a fixed white balance and focus that is set once and left alone. The aim is a sharp, evenly lit, identical frame every time, with black letters and paper that is light without being blown out to pure white.

Getting exposure right for paper

A camera's meter assumes the scene averages out to a middle gray. A sheet of white paper is far brighter than that, so an automatic exposure renders it dull and gray. In manual mode, raise the exposure until the paper looks white on the screen, then check the histogram: the peak for the paper should sit toward the right side without touching the right edge. Touching the edge means faint pencil and light stamps are being erased.

Lens choice

Use a lens of normal or short telephoto length. A wide-angle lens bends straight lines near the edges and forces the camera close to the page, where it shades the paper. If you have a zoom, set it near the middle of its range, and tape the zoom ring if it creeps when pointed down.

Hand-held, when a stand is not allowed

Without a stand you have to trade some quality for a shutter speed fast enough to freeze your hands. Keep the aperture, raise the ISO until the shutter speed is short enough that repeated test frames come out sharp, and brace your elbows on the table. Check sharpness by zooming in on the screen every few dozen frames. How many pixels the page needs, and how far back you can stand, is worked out in the post on how many megapixels a document needs.

RAW, JPEG or TIFF from a camera for OCR

For OCR, shoot JPEG at the camera's highest quality setting, or RAW plus JPEG if you want a master to fall back on. A RAW file is unprocessed sensor data that has to be developed in software before you can upload it as a JPEG or PNG. TIFF is safe and enormous. Your camera's manual says whether it writes one.

Three practical points follow from the choice.

Compression level matters more than format. A JPEG saved at a low quality setting puts blocky halos around letters. Pick the setting your camera labels fine, extra fine or similar, and leave it there.

Check what the next step accepts. Schema Law takes a PDF, PNG or JPEG of up to 20 MB, as the upload step on the homepage states, so RAW and TIFF files have to be exported to one of those first. A high-quality JPEG from a camera with a very large sensor can exceed 20 MB. If yours does, choose the camera's next image size down, after confirming with the megapixel arithmetic that the page still has enough pixels.

Keep the camera originals. Whatever you convert, store the files as they came off the card, unrenamed and unedited, in a folder of their own. If a question comes up about what a page looked like, the untouched frame answers it.

Keeping camera frames in page order across a day of shooting

Camera frames stay in page order when the camera's file numbers run in one unbroken sequence and every box, folder or volume begins with a photographed label, called a slate. The numbers give the order within a folder. The slates tell you which folder you are in when you look at a thousand thumbnails a week later.

Before the first frame

1. Set the camera's clock to the correct local time.
2. Set file numbering to continuous, so the count does not restart when you change cards.
3. Format an empty card in the camera.
4. Prepare a stack of index cards and a marker for slates, if the reading room allows them at the desk. If it allows only pencil, write the slates in pencil.

While shooting

• Slate first. Write the collection, box, folder or volume and the date on a card and photograph it before the first page of each unit. Photograph a card reading "END" after the last page.
• One rule for backs. Decide at the start whether you shoot every reverse side or only those with writing, and keep to it. If you skip blank backs, say so on the slate.
• Never delete in the camera. A bad frame followed by a good one is easy to sort out later. A gap in the numbers looks like a missing page.
• Retake at once. If a frame is soft, shoot the page again immediately so the retake sits next to the failure.
• Keep a paper log. One line per unit: slate text, first frame number, last frame number, anything odd, such as a torn leaf or an enclosure.

A worked example

Suppose a fictional probate dispute over the estate of Harold Brandt turns on entries in three bound ledgers held at a county record office. A paralegal, Ines Carvalho, has one day there. Her log at the end of the day reads:

Back at the office she copies both cards into a folder of originals, then makes working copies renamed by slate and sequence, such as Ledger2_0015.jpg. Because the numbers never restarted and never skipped, she can show that frame 0083 sits where the receipt sat in the book. When the working copies are converted to text, the same order carries through to the output.

What do reading rooms and record offices allow a camera to do?

Reading rooms and record offices set their own camera rules, and those rules differ on the points that shape your rig: whether a tripod or stand may come in, whether flash is allowed, and what you may do with the pictures. Read the institution's published rules and ask the staff before the day you travel.

Two national archives show how far the rules can differ.

The United States National Archives lists equipment in its research room guidance. Under special equipment that is allowed it names "Cameras, camcorders, and tripods," "Photographic copy stands (with approval)" and "Scanners (only contactless or flatbed without auto feed)." Under equipment that is not allowed it names "Flash photography or bulbs," "Scanners with auto feed attachments, whether the attachment is disabled or not" and "Scanning wands or pens." It adds that all equipment must be removed from cases and the cases stored in lockers.

The National Archives of the United Kingdom publishes separate guidance on photographing documents in its reading rooms. It permits imaging with "a hand-held camera, smartphone, tablet, or laptop" and says this is "for personal use only, including research and study." It states that "Flash photography is not permitted," and it lists a tripod among the extendable devices that are not permitted. It also says that "Dedicated camera stands with a standard screw-in mount are available in our reading rooms," with too few for every visitor.

One archive allows your tripod and the other bars it. Both forbid flash. Neither page tells you anything about a county recorder, a court clerk's office or a private collection, each of which has its own practice.

Questions to settle before you go

• May I bring a camera, and does a phone count as one?
• May I bring a tripod or copy stand, or do you provide stands? Do I need to book one?
• Is extra lighting allowed, or must I work by room light?
• May I use slates, tape or weights on the table?
• Is there a limit on the number of pages or a fee per image?
• What may the pictures be used for? If the stated permission is for personal research and your purpose is a court filing, ask what the institution requires.
• If I need a certified copy of a record, how is that ordered?

That last question matters for evidence. A photograph you took yourself and a copy certified by the custodian are different things, and which one your court or agency accepts is a question for its rules. Photograph for reading, searching and working out what to order. For the pages the case turns on, ask how certified copies are ordered.