White glyphs when selecting ocr-text in Evince

[robinrosenstock]

Problem in evince pdf reader:

It only happens when selecting. Is this a display failure? missing fonts? otherwise ocr text is correct.
Similar to #178?

[robinrosenstock]

Found this: https://unix.stackexchange.com/questions/306051/tesseract-is-it-possible-to-change-font-output-in-ocred-pdf

Do I really need to rebuild tesseract for this? Is there no other way around with OCRmyPDF?

[robinrosenstock]

This does only happen with scanned pages?

[jbarlow83]

At a glance it looks like that version of evince can't display the Tesseract glyphless font correctly, although that would be surprising since it's been checked.

Are you using Tesseract 4? What command line?

Try changing the PDF renderer
https://ocrmypdf.readthedocs.io/en/latest/advanced.html#changing-the-pdf-renderer

Try building a regular PDF instead of PDF/A
--output-type pdf

Can you use Tesseract on an image to create a PDF and view that PDF?

[robinrosenstock]

Thanks for your help:
tesseract --version
tesseract 3.05.01
leptonica-1.75.3
libgif 5.1.4 : libjpeg 8d (libjpeg-turbo 1.5.3) : libpng 1.6.34 : libtiff 4.0.9 : zlib 1.2.11 : libwebp 0.6.1

--output-type pdf does not work.
Haven't try to use tesseract on an image directly, because I don't use images only pdfs as input formats.
But when using the hocr renderer it works.
Then I will wait till the arch repos upgrade tesseract to version 4, in hope that other renderers will work, too.
Thanks for you help. I will close this. But if someone has other opinions please go ahead.

[jbarlow83]

What Linux and version of evince? This might be something that the maintainers of Tesseract or evince need to take up.

[robinrosenstock]

Evince Version: 3.26.0+14+g2a499547-1
Linux Kernel version: 4.14.31-1

[amitdo]

tesseract-ocr/tesseract#182 (comment)

[bitwave]

I have the same problem.
evince version: 3.28.2
Linux Kernel version: 4.17.11

looks like the problem lies in ghostscript: tesseract-ocr/tesseract#712

[jbarlow83]

@bitwave The Ghostscript issue is fixed now. An older version of Tesseract/Ghostscript will still be problematic, but I managed to replicate the problem with gs 9.23.

I believe the issue is with Evince itself, so I reported it there.
https://gitlab.gnome.org/GNOME/evince/issues/953

[amitdo]

https://gitlab.freedesktop.org/poppler/poppler/issues/157

[ghost]

Opening the pdf with a web browser allows to see the text while selecting it

[titaniumbones]

I'm still seeing this issue in poppler-based viewers. Is there any workaround available?

[grossherr]

@titaniumbones as mentioned above, change the renderer to hocr, so use --pdf-renderer hocr. Works for me in evince.

[titaniumbones]

where do I add this switch (--pdf-renderer hocr)? In tesseract? evince? just in poppler somehow? I'm not seeing that switch in any docs....

THanks for the help!

[titaniumbones]

Ah shoot of course you meant in ocrmypdf. That is really helpful.

[shinygnu]

Thank you all, this helped me fix what I thought was a deficiency in OCRmyPDF. For me the boxes were black not white, and manifested in all viewers whether Evince, Okular, or PDF-Tools in Emacs.

Adding --pdf-renderer hocr makes highlighted text visible again. Beautiful!

[amitdo]

https://gitlab.freedesktop.org/poppler/poppler/merge_requests/280

@julian-klode, can you please do what the maintainer asked for in your PR?

[julian-klode]

I'd love too but I broke my wrist and the other side elbow, so I'm a bit incapacitated

[amitdo]

Hope you will feel better soon :-)

[amitdo]

https://gitlab.freedesktop.org/poppler/poppler/-/commit/375809c897a66f