Xerox Corporation has posted a warning on their website about an accuracy issue when scanning documents using particular compression levels and resolution settings.
When scanning using the ‘Normal’ quality setting in some Xerox products, the scanned image may be inaccurate. Specifically, digits in numbers may be arbitrarily replaced. Note that this is not an OCR issue – the actual scanned image is incorrect.
What you should do
If you produce scanned documents using a Xerox scanner, you should ensure that you are using the factory defaults with a quality level set to “higher”. Use of the ‘normal’ quality level must be avoided.
PROV advises that any agencies using Xerox scanners should visit the Xerox site listed above for up to date information and advice; check their settings; and contact their support providers if necessary.
The issue was identified on 5 August, when a blogger reported on an apparent bug in some Xerox devices that appeared to incorrectly reproduce characters when scanning. Note: This link is to the original blog post, the blogger has added other posts as further information came to hand.
In the cases reported to date, complete numbers and individual digits have been incorrectly substituted by the scanning process. The bug is particularly problematic as the errors introduced to the scanned document are not immediately apparent, unless a thorough visual check is performed each time.
The issue appears to be with the compression format JBIG2 which is only activated when the operator selects ‘Normal’ scan quality. The higher quality settings use a different format that does not appear to produce the problem.
UPDATE: The blogger who placed the original warning about this issue is receiving reports that the problem may not be restricted to use of the ‘Normal’ setting, and may extend to products by other manufacturers. As these additional claims have not been confirmed at time of this post, PROV recommends that any agencies using scanners make their own enquiries and if necessary take action to reduce their risk.