5

I have hundreds of documents, which are rounded: the document was not straight when it was photographed, and the image is curved or rounded in various ways.

I'm looking for an automatic or semi-automatic process, that will

  • detect the lines (not OCR, just line curves),
  • straighten the whole page according to the (assumed) text lines.

Can you point me a software or feature, or even code that tries to fix this?

example attached (original):

enter image description here

after an (imperfect) manual fix (gimp, cage transformation):

enter image description here

  • if u have only two case (left page and right page) and all pictures shot from same angle(same distort on each page) then i might help you for some action or script for photoshop but still we have many professional for gimp so just wait for their answer :) – Mr.Online Apr 15 '19 at 11:27
  • 1
    It might help to know that this language (Hebrew) "hangs" its letters from a "baseline" that runs across the tops of the letters' x-height rather than English which runs across the bottoms of the letterforms. The Arabic numbers "sit" a conventional baseline that has been imposed upon the average base of the Hebrew line alignment. – Stan Apr 15 '19 at 15:03
  • @stan - Hebrew is exactly like English in this respect: It has a bottom baseline. and like other latins, some letters are taller (לf), some go down (קךgy). – Berry Tsakala May 2 '19 at 12:29
0

Free experimental program Scan Tailor made this:

enter image description here

The shown image used automatic dewarping. It really seems to reduce curvature substantially with no effort.

There is also manual dewarping grid which felt very responsive an gave better results:

enter image description here

enter image description here

The program asks in the beginning which folder contains the page scans that should be straightened and is the writing European or something like Hebrew or Arabic.

The process is semiautomatic, for every page one must select the interesting area, margins and possibly needed rotation & skewing or accept automatic guesses.

I have not tried the program for any big projects, only tested it for separate page scans. It's available here https://scantailor.org

| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.