Xerox devices randomly altering numbers

[nsaspook]

http://www.dkriesel.com/en/blog/201..._are_switching_written_numbers_when_scanning?

http://realbusinessatxerox.blogs.xe...n-on-scanning-issue/?CMP=SMO-EXT#.UgFXIiblAW2

Recently there have been articles about Xerox devices randomly altering numbers in scanned documents. We take this issue very seriously.

It's looks like compression hash collision errors.

 

[DavidSnider]

Holy crap. I would have never imagined a xerox machine could make that sort of error. Their scanning algorithms must be pretty darn clever to even have that sort of bug.

 

[jtbell]

Those eunuch programmers have gotten too clever.

 

[nsaspook]

Those eunuch programmers have gotten too clever.

They were very clever, to the point that saving document space became more important than document image integrity. Lossy compression should never be used when you expect the fidelity of the result to be a limited only by the resolution of the scanning process and the ability of the compression process to eliminate 'redundant' information. It's much better to see a blurry blob than a clear image that's completely wrong in any document scanning operation. Lossy compression usually goes beyond the 'redundant' and starts 'guessing' about different but similar patterns. In this case some numbers are deemed visually similar at some lossy compression ratio so it (the software classifier/comparator) selected a previously stored (seemly random) version to save space.

 

[robphy]

It also seems crazy that Xorex has known about the issue
(it's apparently in its user manual [ http://it.slashdot.org/comments.pl?sid=4058975&cid=44485351 ] )
but made it a feature of the "Normal" option of "copying".

 

[nsaspook]

http://www.cbronline.com/news/tech/hardware/desktops/xerox-to-patch-scan-software-error

They will have a patch to make 'normal' safe again.