[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Scanning?
Eric Smith wrote:
Mike Maginnis <maginnis@tarnover.org> writes:
Don't know if this has been addressed before. I'm planning to re-scan
the Computist magazines that I've already done so far, and get moving
on finishing the rest sometime this decade.
Any thoughts on the best way to scan these? I can't seem to get a
decent image down to less than ~200K. Anything smaller and it image
incurs serious quality loss.
For pages that only contain text and line art, *please* use Group 4 fax
lossless compression. This is a natively supported format in TIFF,
Postscript, and PDF files, and results in 30K-50K sizes for a typical
page (though more for pages with lots of small type or fine drawing
details). Any good scanning software can do it. Since it's lossless,
the quality is identical to the original scan.
This assumes that the original scan was already set up to drop out
the paper grain, ink mottling, etc., etc. Document scanning of the
type we are discussing is _all about_ losing irrelevant details while
preserving a few relevant ones.
However, G4 compression is not good for photographs or other continuous-tone
images, because that's not what it's designed for. Pages that have such
images (other than relatively unimportant incidental ones) should be
compressed with JPEG compression, which is what it sounds like you've
been doing. Unfortunately JPEG compression blurs text and line art.
Ideally, for pages that consist of a mix of text and line art, and
continuous-tone images, you'd use a graphic editor (e.g., Gimp or
Photoshop) to separate the continuous-tone images into a separate layer,
apply JPEG compression to those only, and G4 to the text and line art.
Then you can composite the result in either Postscript or PDF.
That process yields really nice looking results, but takes a lot of
work.
I certainly agree that JPEG is not designed for text. I also
appreciate that a fairly simple, automatable process will be
required if any great number of publications are to be saved.
GIF without error diffusion is also lossless _except for grayscale
quantization_, which is true of any FAX sceme as well.
I suppose it's a matter of taste whether to use 2x the pixels and
1-bit pixels or half as many 2-bit pixels. In most cases, I find
that preserving a couple of midtones helps greatly with anti-aliasing
a scan (aliasing is what makes all FAX scans look like FAXes).
The most critical thing, of course, is dropping out the paper grain,
and, to a lesser extent, the ink noise. If the original was laser
printed, and has lots of tiny toner spots, then despeckling can also
improve compression. The point is to use all the bandwidth for data
instead of noise.
-michael
8-voice music synthesizer using NadaNet networking!
Home page: http://members.aol.com/MJMahon/