[rescue] Risks of lossy algorithms - was Re: Scanning Best Practices - was Re: Inside Solaris newsletter

[Jonathan Patschke]

On Sun, 5 Oct 2014, Toby Thain wrote:

> Perhaps - but I am not sure it's possible to rule this out in any djvu
> implementation. Do you happen to be able to find the original citation?
> I couldn't find it in a few minutes' googling.

There's this:

http://www.dkriesel.com/en/blog/2013/0802_xerox-workcentres_are_switching_written_numbers_when_scanning

And JBIG2 is similar-enough to the JB2 compression algorithm used in DjVu
(both use coalesced references to similar-enough data) that the same sort
of problem is possible, but, generally, it's an implementation problem
rather than an algorithmic one.  Really, the big problem with these sorts
of algorithms that is compression "artifacts" don't call themselves out
because they're copied from other parts of the document rather than
squished to ugliness.

Nearly all the Google results I can find on jbig2 or jb2 altering numbers
terminates in a discussion on the Xerox debacle.

Scanning with sufficient resolution would probably resolve the whole
thing.

-- 
Jonathan Patschke | "Do the difficult things while they are easy and do
Elgin, TX         |  the great things while they are small."
USA               |                                           --Lao Tzu