Scanning

Capture Community Wiki

Scanning is the actual scanning and transformation of paper documents into an electronic image. Can also mean the digitizing of microfilm. In addition to scanners, electronic images can also be captured by:

• Fax – software can read from the fax server. Be aware that image quality will be lower, which may negatively affect recognition accuracy
• Camera phone – Higher-resolution cameras in cell phones, and software designed to work on a cell phone, allow the capture and conversion of documents on the go – whether a conference whiteboard or a restaurant menu.
• Multifunction Device (or Peripheral)– Network-connected MFDs can suffice for low-volume imaging needs.
  

Type of Scanner
Document throughput in the real world will be slower than the scanner’s rated speed. Depending on the volume of paper to be scanned, there’s a scanner to meet the need. Be aware that scanners do not run at their rated speed in the real world. Actual throughput will be less. Plan accordingly.

Scanner Categories:
Workgroup: 10 – 25 ppm
Departmental: 26 – 40 ppm
Mid-volume production 40 – 120 ppm
Production: 120 plus
Large format: for over-sized documents
Check scanners: Will read the MICR number on a check, speeding processing.
Microfilm: for digitizing film-based documents

Scanner features and things to consider:

• Sheetfed or flatbed. For high throughput, a sheetfed scanner with an automatic document feeder is the way to go. For situations with fragile or documents in poor shape, a flatbed scanner may be necessary. Specialty scanners for fragile books/documents as well large formats (engineering drawings, etc.) are also available.
• Resolution. 100 to 1600 dpi. Greater resolution means more storage and bandwidth requirements, even with compression. 300 dpi is good enough for solid recognition results.
• Color, black and white, or grayscale. Grayscale is often the compromise here, as it’s good for OCR and allows graphics to be read. Color is an option, but the improvement in clarity can be offset by the decrease in throughput speed and file size.
• Duplex or simplex. If many double-sided forms, consider purchasing a duplex scanner which scans both sides at once.
• Duty cycle. How long will the scanner last?
• Consumables. Belts and rollers wear out. Are they user replaceable or is a service call required? You don’t want your scanner sitting idle while waiting for a service technician to replace a light bulb.
• Warranty. Should the scanner break, how long will it take to get another?
  

Software Cleanup
Many products include image-enhancement features to increase the quality of the scanned documents – deskew, despeckle, crop, rotate, and/or blank page and double feed detection, etc.

Scanner Drivers
TWAIN
TWAIN may be the most popular of all the interfaces, in part I would guess, due to its early origin (1992) and freedom from royalties. The document is available on the TWAIN working group website (www.twain.org). TWAIN is not an acronym, but originates from a Kipling phrase—“never the twain shall meet”—which reflected the initial developers’ frustrations with making scanners work with personal computers.

The TWAIN Working Group has recently released version 2.0 of the interface specification (about 8 years in the making), which adds a number of capabilities, including support for several Unix/Linux platforms, 64-bit operating environments, check scanning, and additional file formats such as PDF/A, DjVu, and JPEG2000. The Working Group provides a TWAIN software (with the exception of the device-specific logic) open source template, it simplifies development for vendors.

While TWAIN provides both a programmatic and a user interface, most developers call the UI interface. Unfortunately, TWAIN’s design merges the user interface with the device driver. This means that a TWAIN device driver is tied to a graphical environment (See discussions below). Software vendors complain that the differences between implementations complicate support.

ISIS (Image and Scanner Interface Specification)
ISIS was developed by Pixel Translations (now a part of EMC) in 1990. It has a modular, message based structure that helps with multi-processing and performance.

The ISIS specification is openly available, while control and extension of the ISIS specification is controlled by EMC. An earlier version of ISIS was standardized as AIIM MS-61-1996. Either EMC or a vendor may develop an ISIS interface, but EMC requires a vendor to have them test and certify a driver for it to be labeled as “ISIS.” Most are implemented by EMC for a fee, but as a result get support and enhancements when the specification is revised. ISIS application developers purchase their toolkits from EMC and pay a per station fee license to supply it.

Many imaging users believe that ISIS interfaces are faster than TWAIN. It is a generally accepted attitude that production scanners should use ISIS rather than TWAIN. ISIS was early to support enhanced features, such as multi-streaming (scanner supplies multiple image formats from one scan), scan-ahead, and scanner-based barcode recognition.

Home use and low-end scanners often do not come with an ISIS driver, probably due to the fees involved, and the absence of a need for sophisticated features.

WIA (Windows Imaging Architecture also Windows Imaging Acquisition)
The Windows Imaging Architecture is a driver and API model developed by Microsoft that has been extended and refined with successive versions of Windows since 2000. It is an enhancement of the earlier Still Image Architecture in Windows 98. At one time WIA supported video, but that has been moved to a separate API with the introduction of Vista.

Unlike TWAIN (and like ISIS), WIA separates driver activity from user interface, running them in different processor address spaces for better security and reliability. WIA enables Microsoft operating systems to perform plug-and-play recognition of scanners, and support multiple scanners on a system (as do some of the other interfaces). Some developers feel that it is a simpler interface to build than TWAIN, perhaps because Microsoft provides a library that performs most driver functions. Even with that, there don’t seem to be that many WIA drivers so far.

Of course, WIA is only available on Windows systems, and seems to be an evolving target. One has to realize that its evolution has been and will be controlled by Microsoft’s goals, and not scanner capabilities or developer needs. But at least there aren’t any license fees!

SANE (Scanner Access Now Easy)
In my research I discovered yet another scanning interface, SANE. SANE is a public domain open development API designed to provide access to raster imaging hardware. Originally targeted at Unix/Linux, it is now available on any platform. Like ISIS and WIA, SANE eliminates the user interface, making it easier to have one driver support multiple software packages.

I must admit that I have never used a SANE driver, and I don’t know of any commercial scanning packages that support it; everything listed on the SANE website (http://www.sane-project.org) is open source. I’ve scheduled a project to try installing and using SANE with my lab scanners sometime this summer. Look here in a few months for more data.

Making the Choice
Well, say I just bought a new scanner, and it has TWAIN, ISIS, and WIA support. What do I choose? Well, I have to admit a personal bias for ISIS when it is available. ISIS is tried and true, and tends to keep up with new scanner capabilities as they appear. There is no question that an ISIS interface will do its best to run at rated speeds in most situations.

In the end, what matters is what works. If one interface is cumbersome or doesn’t work well, try another. Choice is a good thing.

Scanner driver information originally written by Bernard Chester, IMERGE Consulting, for the May/June 2009 issue of Infonomics magazine.