Skew is annoying when you’re taking photos, but it can be a real problem when you’re scanning a document and trying to read what’s on it. A look at the problem and various methods for correcting it.
If you’ve ever done any job that involved taking pictures – whether that means with a camera, a scanner, or anything else – you’ve almost certainly experienced the issue of image alignment and skew. Whether you call it by that name or something else, it’s the cause of countless photos and documents that are tilted, cut off, unreadable, or otherwise unusable.
The Big 3 Causes of Image Skew
In simplest terms, image skew means that your image is crooked. How it got that way is usually a particular blend of man, machine, and user error.
Unsteady camera – If you’re taking a picture of a stationary object, there’s really only one way to get skew, and that’s if the camera isn’t level relative to the object.
Misaligned document – On a flatbed scanner, even though both the camera and the document are in fixed positions, you can get a skewed image if you didn’t lay the paper straight, or if you disturbed it when you closed the lid.
Feeding problems – In machine-fed scanners, skew can happen when one edge of an item inserted into the feeder isn’t lying flat horizontally. This can be exacerbated by the action of the rollers as they attempt to feed the item through the track.
For cameras, the solution for over a century was pretty straightforward: Get a tripod, or some other mechanism that can hold the shot steadier than your hand can. In the digital age, it’s become possible to compensate for an unsteady picture with image stabilization technology, both hardware- and software-based. We won’t get into too much detail here, but the software-based kind basically works by taking a wider picture than necessary and leaving extra room to tilt the image back to level. Hardware-based stabilization actually senses movement in the camera and adjusts the optics.
Scanners similarly have both hardware- and software-based anti-skew technology, which we’ll talk about in a minute. We’ve developed many of these over a long history of working with cameras and optics, long before desktop check scanners or ID scanners were invented!
Image Skew: An Ancient History
Some of our own company’s earliest experiences with image skew came more than 50 years ago during the golden age of space exploration in the 1970s. Back then, NASA hired our California manufacturing division (which at the time went by the name Electro-Optical Mechanics, or EOM) to make the image conversion cameras for the Mariner and Viking missions to the planet Mars.
That camera had to hold a 5-inch piece of film steady to within 0.0004 inches – a skew tolerance of 0.0091 degrees – as separate red, green, and blue pixels were flashed in sequence on a special TV screen, turning magnetic reel data from the probes into visible photos. Any movement at all would blur the image. The engineers devised a clever solution: They cut invisible hairline grooves in the back of the glass that the film pressed against behind the lens, and sucked the air out through the grooves with a vacuum pump, holding the film perfectly still. A purely analog solution to image skew, but it was responsible for many if not most of the color photographs of Mars and Venus that you’ve seen from those early missions!
How we solve image skew today
While the NASA project is a neat backstory, keeping a piece of film motionless is a lot different from keeping a moving check or ID card perfectly straight as it goes past a camera at high speed.
The first line of defense against image skew is similar to that used for stabilization by regular cameras: The area captured in the image is bigger than the document itself. This is mostly the result of the natural tolerance built in so that our scanners can accommodate checks of all shapes and sizes.
Similarly, the MICR reader in a check scanner reads a strip about twice as wide as a standard E13B or CMC7 character. That’s to account for variations in the height of different checks, but even with the extra room, the document needs to be very close to perfectly straight in order for the reader to pick up the correct signal strength (see diagram at right).
In both cases, Digital Check employs a patented anti-skew technology that attempts to identify the center of each character and interpret the information in an area about one and a half times the typical character size around it. While this can’t fix magnetic distortion, it is possible to get a visual read on the characters using Optical Character Recognition (OCR) and compare them to the magnetic signal in order to catch errors.
As far as the other big component of image skew goes – that is, feeding the document correctly – it mainly comes down to a combination of pressure and grip strength by the rollers that are pulling it through. That’s why, for example, we added a mechanical plunger to the ID card slot on out check scanners, which effectively forces the card to be inserted straight and holds it there. Other factors like the roller material and shape, as well as the power from the motor, are all adjusted to give the right amount of force for the type of item being inserted. This comes from many years of working with different document types and figuring out what works best.
And of course, the last way we fight image skew is by encouraging good user behavior. Sometimes that means educating the user about the right way to insert a check or an ID card – for example, with the article you’re reading right now, or in the user instructions, or even by markings on the device itself. We also use physical guides that make it easy to insert an item the correct way, but more difficult to do the wrong way.
There are some differences between how an item is handled in a combined check scanner/ID scanner compared to a single-purpose device, which we’ll get into in a separate article about feed mechanisms. For now, we hope you’ve found this an interesting look at an under-the-radar issue with mechanical scanning – and we hope that all your images come out straight!
