need the use of a scanner with sheet feeder (near Seattle)

From: Douglas H. Quebbeman <dquebbeman_at_acm.org>
Date: Mon Apr 8 15:56:30 2002

> > However... have you checked out the DjVu imaging compression
> > technology? DjVu is a non-propreitary superset of the iterated fractal
> > system imaging compression technology I read about in Byte magazine
> > back in the 1980s. Now that the secret's out of the bag, everyone can
> > have utilize extremely high-compression if you can suffer the slight
> > loss of fidelity to the original (for example, the analysis will find
> > a single ideal letter form for an 'A', and uses that ideal letterform
> > image when reconstructing the document, instead of recording every
> > pixel at every location.
>
> Sounds like OCR software to me. There any major differences?

OCR maps letter forms to text; this is pure compression.

Looking at it another way, DjVu condenses any image fed into it
into a mathemetical expression that, when evaluated, yields
as its result, the image of the original document.

So, it's nothing like OCR. If the original image were a page full
of little apples, the program will decide which apple is the best
one, and when it reconstitutes the original image, will put as
many copies of the one apple on the page as the original had. If
there are subtle differences between the apples that the eye
won't readily see, then the reconstituted image won't have those
subtle differences.

It goes beyond this too; it separates the text and calls that
foreground, and everything that's not text is background. The
background is compressed with a different family of wavelets
than is used for the foreground.

It's worth spending an hour googling for better answers than
I could ever hope to provide...

-dq

-Douglas Hurst Quebbeman (DougQ at ixnayamspayIgLou.com) [Call me "Doug"]
  Surgically excise the pig-latin from my e-mail address in order to reply
  "The large print giveth, and the small print taketh away." -Tom Waits
Received on Mon Apr 08 2002 - 15:56:30 BST