get_textpage_ocr() should be able to identify image / page rotation and read the content accordingly

[lmocsi]

Is your feature request related to a problem? Please describe.
There are pdfs where some pages are 90 degrees rotated to the left or right, or upside down rotated. These pages cannot be read by get_textpage_ocr() - I only get garbage as output.

Describe the solution you'd like
What I'd like is that get_textpage_ocr() identifies if a page / image is rotated 90 degrees to the left or right (or upside down rotated), rotated it back an read the content of the upright standing page / image. As of pymupdf==1.27.2.3 it just tries to read the content of the page as is, so from rotated pages it just reads garbage.

Describe alternatives you've considered
Textract on AWS does this automatically.

Additional context

[JorjMcKie]

Have you tried page.remove_rotation()? Please provide an example page.

[lmocsi]

get_textpage_ocr() easily reads the first file.
Fails with the others (rotated Left, Right and UPside down).
If I try page.remove_rotation() I get AttributeError: 'IdentityMatrix' object has no attribute 'get_textpage_ocr' error.

[MannXo]

This looks like something I can tackle, mind assigning me?

[JorjMcKie]

This looks like something I can tackle, mind assigning me?

You are welcome. But I am not convinced yet, that we all have the same understanding of the problem.

1. If the page itself is rotated, then page.remove_rotation() will modify it to have rotation=0 without impacting visibility. After doing this, page.get_textpage_ocr() should work as desired.
2. However: If page rotation is 0 but the full-page image itself is rotated, then how are we supposed to find out? Depending on the type of image, EXIF info might be available to tell us. This would be more effort.
3. Entirely outside scope are other cases: multiple images with possibly different rotation.