Page 8 of 9

Re: Scan Tailor "Enhanced"

Posted: 02 Aug 2012, 19:29
by dtic
Pejuko, do you have any plans for enhancement of the Split Pages step?

Here is a problem scenario when using the middle manual layout mode (split one page from a one-plus-part-of-another-page image).
1.png (3.8 KiB) Viewed 20494 times
ST then sometimes splits in a way that gives an "overwide" (too wide) blue area with unwanted parts on the right, left or both sides included in the blue area. That is a problem especially if a part of the the opposite page is included and there is text reflection in the glass over that part. Some ST projects get a lot of such overwide splits. The manual fix involves finding the problem page, waitingfor the large image view to load, clicking the split line and dragging and dropping it. Repeating that for 100 pages in a large project takes time. An external workaround is to crop the images prior to ST or to change the camera position to include less of the opposite page. But that is not always possible e.g. when using a handheld platen the cradle (and book) might move horizontally during image capture. To capture all of all of the pages the camera must then be zoomed out with a safety margin. And because of the moving cradle there will be no no simple way to batch crop all the images without losing content.

Here follows a sketch of two actions for better handling of such problem cases in ST. I don't know how complex an enhancement that would be to make, maybe you can say something about that? But I post this mostly for discussion. Is there some other better way to deal with the problem? Perhaps some way already available in ST?

1. add "order by increasing width" in the drop down at the bottom in the right hand pane for the Split Pages step. I.e. similar to what we already have for the Select Content step. That would group problem cases (overwide/underwide splits) which saves time compared to finding them one at a time.
2. add two "Refine split" actions. The idea is to take one of the previous two split lines as a baseline and search for a new (next best) split line to the right or left of the baseline. There are four possible such actions. But the two actions for moving a splitline inwards ("Move left side split line to the right" and "move right side split line to the left") would be enough to handle overwide split cases.
If the refine split actions where also put in the thumbnail pane context menu then users could do "order by width" to sort the problem cases together, selects them and then right click one and pick one of the two actions to run on all of them.

Re: Scan Tailor "Enhanced"

Posted: 08 Aug 2012, 23:46
by rdx
pejuko wrote:I've just released new pre version. all deps including qt was built on one machine. hope it helps.
Hi pejuko,

I just want to confirm that the new version opens, but it seems to break a function in the original, non-enhanced version.
When I manually select the content box and attempt to apply it to every other page, the program does not really apply it. When I click on those supposedly-applied pages nothing happens. It works find in the non-enhanced version.

Again, thank you guys a lot for a great piece of software.


Re: Scan Tailor "Enhanced"

Posted: 09 Aug 2012, 03:57
by rdx
Sorry, it seems to work now. I can't reproduce the problem and it may well be something I've done wrongly.

Re: Scan Tailor "Enhanced"

Posted: 06 Oct 2012, 10:29
by eL_PuSHeR
Good day.

Do you intend to release a x64 version in the near future?

Re: Scan Tailor "Enhanced"

Posted: 18 Oct 2012, 22:25
by eldridge
Hey everyone,

I am fairly new to the family here. But, I have been reading the forum for a few months. I recently got my scanner kit and its assembled. For a long while I was doing a destructive scan before I learned about the DIY community. I've used a fujitsu scan snap S1500M. The quality is incomparable to anything else after I run it through my postprocessing procedure in photoshop. However, I am now struggling with learning how to fiddle with ST since using a camera is far different than the fujitsu.

Difficulties I've been having:
1) I'm using Mac OS. Not much support around here in that area that seem to be doing some of the ground breaking things like OpenCV.
2) When I save with ST, it downgrades the quality of the images so dramatically compared to the large photoshop files I'll usually plug into acrobat to OCR and finalize (as I did with my destructive scanning processes).

New Routes Recently Traversed:
1) I have been following the advice of users and I've upgraded to ST Enhanced today through HomeBrew. Using command lines and terminal is somewhat foreign to me. But, the build and install was a success. So thank you to everyone who has been putting in the hours of work here and especially those who contribute to the Mac folk.

Here is my question today:
Using ScanTailor Enhanced, is it possible for me to only have the program go through steps 1-5 and then save those images without executing any of its typical output elements (viz., saving the file in the same format with the same compression type as was inputted)?


Re: Scan Tailor "Enhanced"

Posted: 19 Jan 2013, 06:39
by dtic
Pejuko, I noticed that there has been a number of updates to ST enhanced at sourceforge here ... /enhanced/

Your latest release does not have the "pre" tag. Do you have a changelog for it? the readme at the bottom of the linked page is from May 2012.

Re: Scan Tailor "Enhanced"

Posted: 20 Feb 2013, 06:13
by sidney
o3h1p wrote:Both Pictureshape algorithms yield worse results (both the 'free' and 'rectangular') for me. Same exact page and the 'free' algorithm incorrectly identifies text as a 'picture zone', the 'rectangular' algorithm identifies the entire page as a 'picture zone' but original version 0.9.10 from Talon's website identifies properly (aside from minor issues).
I have the same problem (however used as original version and latest version of scantailor-enhanced).

Re: Scan Tailor "Enhanced"

Posted: 26 Mar 2013, 10:23
by marcusdeman

I've been using Scan Tailor and the Scan Tailor "Enhanced" for some time and I have some questions / ideas.

1. Would it be possible to have the output saved after each stage instead of only at the output stage? Even though ST is awesome
some apps are better with certain stages.

2. Would it be possible to have some preliminary dewarping taking place based on the warping of the outer edges of a page? Perhaps
even at the split pages stages.

Now the dewarping is done on last stage based on text but this is only succesfull when you have one chunk of text. Most of the
times the dewarping goes crazy because of the page numbering.

If this is possible it will also helps in selecting the right size of the content box at the content stage.

Perhaps another way to overcome this is to being able to select a box size where the algorithm only focuses on instead of the whole page. Similar to the focus point (spot/center/whole frame) with digital camera's.

Re: Scan Tailor "Enhanced"

Posted: 19 May 2013, 20:31
by pablitoclavito
Hi! Thanks for bringing new updates of this great program!
sidney wrote:
o3h1p wrote:Both Pictureshape algorithms yield worse results (both the 'free' and 'rectangular') for me. Same exact page and the 'free' algorithm incorrectly identifies text as a 'picture zone', the 'rectangular' algorithm identifies the entire page as a 'picture zone' but original version 0.9.10 from Talon's website identifies properly (aside from minor issues).
I have the same problem (however used as original version and latest version of scantailor-enhanced).
I have the same problem. Text being recognised as image (in the "normal" Scan Tailor version the same photo is recognized as text only).

I am using the last version: scantailor-enhanced-20130508-32bit in Windows 7 x64, from the cli.

Any idea or option to select in the cli to avoid this?
Thanks in advance.

Re: Scan Tailor "Enhanced"

Posted: 19 Apr 2014, 19:53
by dtic
A bug in scantailor-cli.exe : if the switch --dpi is included but no dpi value is set (that is, the command line string includes "--dpi= ") the ST process consumes over 1 gigabyte and takes a very long time compared to if a dpi value, e.g. 200 or 300 or 400, was set. If the dpi value is missing ST should either use e.g. 300 as a default (and make that clear in the help file) or stop with an error. This bug is present in the latest release scantailor-enhanced-20140214-32bit-install and in older releases.

I don't have a github account so anyone with such an account that reads this could help out by adding the above as a bug on the scan tailor github page.