Page 1 of 28

Scan Tailor "Experimental"

Posted: 20 Jun 2015, 13:06
by Tulon
Hi,

You probably still remember me as the original author of Scan Tailor. I stopped being involved in the project a while ago. Prior to that I was working on a major architectural change that would:
* Move dewarping to "Deskew" stage.
* Don't rely on DPIs anymore. In fact, get rid of them altogether.

I was roughly half way there when I lost interest. Some time ago I felt like doing some coding in C++11, which I don't have a chance to use at work, so I ended up working on Scan Tailor again. Eventually I finished the major architectural change, achieving both of the goals mentioned above. It doesn't mean my work is in a releasable state though. Most notably, the command line mode is broken.

Still, I thought it would be a good idea to make a build available for people to try out. So, here it is: https://github.com/Tulon/scantailor/rel ... 2015_06_20

P.S.
I removed the message on proper usage of DPIs from my signature. With Scan Tailor "Experimental" you don't get a chance to mess that up :)

Re: Scan Tailor "Experimental"

Posted: 20 Jun 2015, 19:27
by dingodog
Welcome back Tulon! ;-)

You are unforgettable author of scantailor!

I'm very curious to test this build

Re: Scan Tailor "Experimental"

Posted: 21 Jun 2015, 16:15
by reproman
Good to hear from you.
Your program has been a great help to me.
I have scanned/processed to pdf over 1200 Bibles.
Is it necessary to uninstall my existing 64bit program on windows7 before installing the 32bit experimental?
Thanks,
Jim

Re: Scan Tailor "Experimental"

Posted: 21 Jun 2015, 19:17
by Tulon
Is it necessary to uninstall my existing 64bit program on windows7 before installing the 32bit experimental?
Not strictly necessary, though the new installation will overwrite at least the start menu item, so it's a good idea to re-install your original version afterwards.

Also, be careful not to overwrite your existing projects, as project file format has changed. This experimental version can kind of open the old project files, except most of the settings get lost. The same is true when going the other way.

Re: Scan Tailor "Experimental"

Posted: 23 Jun 2015, 03:08
by reproman
Processed one 1700 page and one 850 page bible earlier today with new experimental 32bit program. Worked fine. Hope you do a 64bit.
Thanks,
j

Re: Scan Tailor "Experimental"

Posted: 23 Jun 2015, 06:57
by Tulon
Processed one 1700 page and one 850 page bible earlier today with new experimental 32bit program. Worked fine.
I guess the workflow didn't change much if you are not using dewarping. Anyone wants to try that and share their experience?
BTW, there is now a dedicated de-keystonong mode.

Re: Scan Tailor "Experimental"

Posted: 28 Jun 2015, 15:50
by b0bcat
Does your 'experimental' build eradicate the "aura" phenomenon?
http://www.diybookscanner.org/forum/vie ... =21&t=3029
I know in that thread it's stated as fixed in the then uncompiled code....
That was the only substantive bug I can recall in ST and I've been waiting for the current maintainer to come to a point where he (or someone) feels like issuing a w32 binary including the fix as I have a backlog of 'aura' afflicted projects.
btw it would be exceptionally useful if your 'experimental' build could be extended to read and write existing format .scantailor project files. My interest in this suggestion is again purely objective (academic) in nature. ;)


(PS the backlog mainly results from scanning old cheaply printed (on absorbent paper) books ca.1900-1930 in b&w 300dpi. These are replete with ink dots which to avoid too much OCR confusion I have manually removed by fill-in of whitespace. A monumentally tiresome and time-consuming task. More recently I have done an original scan in greyscale 300dpi and saved as b&w 300dpi in ST but not noticed any difference in the level of printer's ink artefacts being preserved.)

Re: Scan Tailor "Experimental"

Posted: 28 Jun 2015, 16:52
by Tulon
Does your 'experimental' build eradicate the "aura" phenomenon?
It doesn't and neither does the latest master branch.

The fix for that is non-trivial because it's not a bug but a limitation of the illumination normalization algorithm used. It doesn't like pictures that go up to the very edge of a page. I haven't seen your original scans but I suspect that's the case.

Re: Scan Tailor "Experimental"

Posted: 09 Jul 2015, 15:20
by b0bcat
Thanks for that and the interesting remark about image next to page edge which inspires me to try repositioning the picture zone part of the image in a pre-process step so it has whitespace between it and edge of image, in case that does anything.
My comment above concerning a code fix was in error, I was thinking of the page-linking fix, sorry for any potential confusion.

Re: Scan Tailor "Experimental"

Posted: 14 Jul 2015, 04:10
by xerum
Thank You Tulon for ST Experimental.

It works well however I am not sure if it is just me but the "Page Layout" section when selecting "apply to all pages" does not seem to work or take effect. Neither does the "Apply to all pages below".

I've got all variants of ST installed on my system (in different folders though) so not sure if this is causing glitches.

Can someone tell me if it is just me or they too have the same glitch.

Thanks
Xe