Page 2 of 3

Re: B/W source - how to keep format?

Posted: 05 Oct 2011, 15:49
by eL_PuSHeR
In the interim I have made a 4bpp version that will suffice for now. Thank you very much for your replies. Still waiting Tulon's opinion.

Re: B/W source - how to keep format?

Posted: 06 Oct 2011, 04:37
by Tulon
eL_PuSHeR wrote:Still waiting Tulon's opinion.
I second pejuko's opinion - go with the Mixed mode.
I personally wouldn't care about the output size, as I would convert it to PDF or DJVU and then discard it.

Re: B/W source - how to keep format?

Posted: 06 Oct 2011, 09:07
by eL_PuSHeR
Tried mixed mode today but I don't understand what it does. A new button appears telling something about shape (free, rectangular). I tested it on a page (free) and it looked normal. What's the difference between mixed mode and grayscale mode for instance?

Re: B/W source - how to keep format?

Posted: 06 Oct 2011, 10:15
by Misty
Mixed mode will convert part of the image into pure black/white, and part of it into a greyscale or colour image with multiple shades. The advantage is that pure black/white text compresses much, much better than greyscale. If you use mixed mode, you can then use a program that will separate the two parts and use the optimal compression for each of them; that will get you much better compression than if you did the whole thing greyscale.

By default, Scan Tailor automatically detects where the greyscale or colour parts of the image are, and treats everything else as text. The shape tool lets you manually override that to define certain parts of the image as either being text or greyscale.

Re: B/W source - how to keep format?

Posted: 06 Oct 2011, 15:53
by eL_PuSHeR
Interesting. And how Scan Tailor decides which parts of a bitmap is text and which parts aren't?

Thank you very much for your replies all of you, honest.

Re: B/W source - how to keep format?

Posted: 06 Oct 2011, 16:11
by Misty
It does a visual analysis to try to pick out dark text on light background from illustrations. It's usually quite accurate, but does make mistakes.

Re: B/W source - how to keep format?

Posted: 07 Oct 2011, 02:25
by eL_PuSHeR
Thanks. It seems interesting for scanning text only using line art mode (600dpi-2bpp).

Re: B/W source - how to keep format?

Posted: 07 Oct 2011, 10:49
by eL_PuSHeR
I have made several tests more. I have found out that the way ST treats 2bpp source images defeates most of the reason about using 600dpi line-art mode. I don't know why ST cannot work with pure b/w 2bpp images but after processing some pages I get heavy artifacting no matter what output mode I use. B/W should preserve current pixels/palette but it doesn't. The other two remaining modes produce heavy artifacts clearly noticeable on output specially on dark areas, like rhombus-shaped patterns.

Any ideas are welcome by the way...

Re: B/W source - how to keep format?

Posted: 08 Oct 2011, 15:31
by daniel_reetz
Do you have a bitmap we can look at?

Re: B/W source - how to keep format?

Posted: 08 Oct 2011, 16:20
by dingodog
if eL_Pusher had not afraid to use command-line programs, he might get better results than with scantailor

using jbig2enc

I blurred image with radious 4 before to encode with jbig2enc, as I wrote here

http://www.diybookscanner.org/forum/vie ... 1395#11395

this is a typical fear induced by Micro$oft to avoid that users become too smart