3 Sep
2006
3 Sep
'06
9:50 a.m.
Don wrote:
I had earlier tried some gr[ea]y scale scans and
convinced
myself I must be doing something wrong (as the sizes were
just so much larger) so I didn't pursue them.
Hmm, a greyscale TIFF test would be the most important one I'd say - something
like 16 levels; 256 *might* be overkill.
What's readable to a human is important - but equally so is encoding enough
information such that scans of sub-standard source material (dirty, torn etc.)
could be post-processed on a per-case basis if needs be, before passing to a
subsequent OCR step. This is the bit where bi-level scanning tends to fall
down as it depends where the level between white/black is as to what gets
encoded for "damaged" sections of a document. On pristine source material and
at a high enough resolution (so that viewing on a screen scaled-down gets rid
of the jagged edges) I'm sure it's otherwise fine.
[N.B. the gr[ea]yscale scans are much
"softer" on the eyes
(no doubt due to the continuum of "value")]
Lots of the stuff I come across tends to have at least one or two
continuous-tone photos inside, which if I was scanning at bi-level (which I
never do) I'd have to treat as specific cases.