OSError when accessing images

[elisabethzinck]

I am trying to extract text from a pdf, where I first try extracting the text using extract_text(). If that fails, I want to get the image(s) so that I can extract the text using OCR-technology.

Environment

Which environment were you using when you encountered the problem?

$ python -m platform
macOS-13.2.1-arm64-arm-64bit

$ python -c "import pypdf;print(pypdf.__version__)"
3.8.0

I have also installed Pillow==9.4.0.

Code + PDF

This is a minimal, complete example that shows the issue:

from pypdf import PdfReader

file_path = 'test_img.pdf'
reader = PdfReader(file_path)

page0 = reader.pages[0]
text = page0.extract_text()

if text.strip() == '':
    page0.images

The pdf used in the example is included below:
test_img.pdf

Traceback

This is the complete Traceback I see:

Traceback (most recent call last):
  File "/home/moose/.pyenv/versions/3.11.1/lib/python3.11/site-packages/PIL/PngImagePlugin.py", line 1286, in _save
    rawmode, mode = _OUTMODES[mode]
                    ~~~~~~~~~^^^^^^
KeyError: 'PA'

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "<stdin>", line 2, in <module>
  File "/home/moose/.pyenv/versions/3.11.1/lib/python3.11/site-packages/pypdf/_page.py", line 444, in images
    extension, byte_stream = _xobj_to_image(x_object[obj])
                             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/moose/.pyenv/versions/3.11.1/lib/python3.11/site-packages/pypdf/filters.py", line 690, in _xobj_to_image
    img.save(img_byte_arr, format="PNG")
  File "/home/moose/.pyenv/versions/3.11.1/lib/python3.11/site-packages/PIL/Image.py", line 2432, in save
    save_handler(self, fp, filename)
  File "/home/moose/.pyenv/versions/3.11.1/lib/python3.11/site-packages/PIL/PngImagePlugin.py", line 1289, in _save
    raise OSError(msg) from e
OSError: cannot write mode PA as PNG

Is there any way the format can be determined in pypdf so we don't get the error from PIL? Or is there a way around the error?

[MartinThoma]

Thank you for adding such a good issue description :-)

I think I found & fixed the problem in #1802 . Would you mind to check if that solves the issue for you?

[MartinThoma]

I'm a bit confused as the extracted image does not look like the PDF. So my fix is likely flawed. Does somebody have an idea what the issue is?

[elisabethzinck]

Glad you liked my issue description - it's the first issue I've opened :)

Yes, when installing the issue-1801 branch version, I don't get the error anymore. But yes, in the output image the text "Test" is written with some strange horizontal lines.

[MartinThoma]

I think I made it work :-) Here is the image:

Is it ok if I add test_img.pdf to https://github.com/py-pdf/sample-files ?

[elisabethzinck]

Awesome - it works perfectly now! Thank you for your help.

Sure! :)

[MartinThoma]

The fix was just merged to main and will be released on Sunday with pypdf > 3.8.0

[MartinThoma]

Thank you for your help!

If you want, I can add you as a contributor: https://pypdf.readthedocs.io/en/latest/meta/CONTRIBUTORS.html

[elisabethzinck]

Great - looking forward to use the new version.

Thanks for the offer, but no thanks. :)