question about training tesseract

[morteza neishaboori]

Hello,

I want to train tesseract to detect words in such images in the link below!

https://drive.google.com/folderview?id=0B3dLM0w0EeD-RFZVc1NjaGNqUlE&usp=sharing

I tried but it was not successful! now I will be happy if somebody can give me some hints if it's at all possible to do this with tesseract?!

[Paul]

This paper suggests a binarization approach that might be helpful with your imagery. Unfortunately you need to implement it on your own in a preprocessing step,

since Tesseract only uses Otsu's method for binarization. Thus the bad results.

[zdenko podobny]

Well, leptonica also provide some binarization methods (see source code[1]). Some explanation can be found at web[2]. Of course there are other binarization methods with published code - e.g. c++ source code for Niblack, Wolf can be found on christian wolf page[3]

IMO in this case it should be worthy to have a look at page segmentation - there is (older) presentation of leptonica posibilities[4].

BTW: It looks like jasonlfunk made an implementation of Kasar, Kumar, Ramakrishnan paper using python and opencv[5]....

[1] https://tpgit.github.io/Leptonica/binarize_8c.html

[2] http://www.leptonica.com/binarization.html

[3] http://liris.cnrs.fr/christian.wolf/software/binarize/

[4] http://www.dicklyon.com/phototech/PhotoTech_11_DocImage_Slides.pdf

[5] https://github.com/jasonlfunk/ocr-text-extraction

Zdenko

--
You received this message because you are subscribed to the Google Groups "tesseract-ocr" group.
To unsubscribe from this group and stop receiving emails from it, send an email to tesseract-oc...@googlegroups.com.
To post to this group, send email to tesser...@googlegroups.com.
Visit this group at http://groups.google.com/group/tesseract-ocr.

To view this discussion on the web visit https://groups.google.com/d/msgid/tesseract-ocr/c149a5a9-f72c-4fa8-8f78-9432715d380c%40googlegroups.com.

For more options, visit https://groups.google.com/d/optout.