25 Sep
2015
25 Sep
'15
11:06 a.m.
On 2015-09-25 16:15, Paul Koning wrote:
Agreed that it always needs checking/cleanup. But I seriously would like
to OCR the manuals, instead of dealing bit bitmaps.
Johnny
On Sep 24, 2015, at 6:12 PM, Johnny Billquist
<bqt at Update.UU.SE> wrote:
OCR is fine but only if the bitmap image is available, OR if the resulting text is
thoroughly cleaned up. No OCR program can produce accurate text from scanned documents,
even if the scans are good quality. And not all scans are. ...
I'm getting reasonable results with 4bpp indexed color .pngs
Several RSX manuals have important color coding. Some manuals have both black, red and
blue text. Others have sections with a light red or gray background.
I have paper versions of them. Unfortunately, all scanned manuals I've seen have been
plain b/w. Also, bitmaps, and not OCRed. :-(