29 Nov
2006
29 Nov
'06
1:41 p.m.
Jules Richardson wrote:
It's OK if you have top-quality documentation. But
lots of computer
docs out there are old, faded, dirty, creased, well-thumbed etc. and
unless someone's prepared to visually check every scanned page,
there's a chance that the bi-level algorithm in use will corrupt the
data and it'll go unnoticed.
Well I do look at each page in the final scan - pretty briefly just
to make sure that all the pages are in the correct order. Even so
I've never noticed a problem. I've never tried scanning a line-printer
listing and most stuff I scan is a reasonable manual, but even on
the few scans I've made of photocopies, I've never seen a problem.
It would have to be a gross problem to affect legibility of text
by a human (almost anything affects OCR :-( ).
I suppose it's one of those situations where you
end up throwing away
information no matter what (after all, any scan is essentially a
digital representation of analogue data), but there's a danger of
throwing away too much data - and for rare docs you might only get
the chance to scan them once. For rare items I'd rather have maximum
quality "just in case", even if it does mean more storage space.
I scan at 600dpi bi-level G4-encoded and I end up with some manuals
in the hundreds of MB. I think at least one of the Digital Semi
manuals hits nearly 400MB. If I went to greyscale (8 bits) then
some of those manuals would take half a DVD[1] to store. Perhaps
a few years from now I'll come back and look for some way of
erasing this post ("2GB, that's nothing, I can downlaod that
in 10 mins!!") but right now 300MB is manageable (even as a download)
whereas 2GB is a stretch.
For my first few manuals I did redo pages in greyscale if they
had a photo but I stopped because I could not see a difference.
But by all means scan in 8-bits, just don't scan text as a JPEG
unless the alternative is not scanning at all :-)
Antonio
[1] Maybe more - I think G4 only works on bi-level, so by using
greyscale I think you lose some lossless compression too.