I tried it on a page from the scans made by Sean. I used Finereader PRO 5,
one of the best OCR programs around, i am used to 100% errorfree ocr results
with this program.
Not on this scan, the error rate was quite high. Too low a resolution (300
dpi is adequate) and it is is a low quality color scan and not a high
contrast black and white.
When the ocr program converted the picture to black and white i could even
see fingerprints. With this original using ocr will be a task.
But with a better quality scan (by preference directly from the ocr program)
it can be done in a reasonable tiem, ocr programs are quie good in not only
chracter recognition but also keeping the layout intact.
Not that i am ungrateful, far from that. I appreciate very much the work put
in the scans by Sean.
-----Original Message-----
From: Ricardo Jurczyk Pinheiro [mailto:[EMAIL PROTECTED]]
Sent: 21 February 2001 04:07
To: [EMAIL PROTECTED]
Subject: Re: v9938 databook scanned
Em sex, 16 fev 2001, Sean Young escreveu:
> > > JPG is meant for photographs and nothing else. (Stuff with smooth
color
> > > gradients.)
> >
> > Why not 256 color (or even B&W) GIFs?
>
> GIFs are evil. From http://www.libpng.org/pub/png/ :
(...)
Sorry, I forgot that "little problem" with GIFs. =)
Well, I think it'd be interesting to use an OCR program to
convert all PNG files to text.
--------------------------------------------------------------------------
Ricardo Jurczyk Pinheiro - M. Sc. Numerical Modelling - [EMAIL PROTECTED] - 3635907
[EMAIL PROTECTED] - Anime, ABU, MSX, Linux, Gospel, ST, Rock, Math
Sola Scriptura - Sola Gratia - Sola Fide - Solo Christi - Soli Deo Gloria
Time heals all wounds, but the belly button remains.
--
For info, see http://www.stack.nl/~wynke/MSX/listinfo.html
--
For info, see http://www.stack.nl/~wynke/MSX/listinfo.html
