PageObject.images compresses JPEG images with only 75% quality, and there is no option to control this

[Lucas-C]

Hi!

I'm a maintainer of pdfly, that has an extract-images subcommand, that uses PageObject.images internally.

Recently, a user reported an image compression issue: py-pdf/pdfly#200

I investigated and I think I found what happens:

• in pypdf, PageObject.images invokes PageObject._get_image() that calls _xobj_to_image()
• _xobj_to_image() calls the Image.save() method of Pillow that use a compression of 75% by default: https://pillow.readthedocs.io/en/stable/handbook/image-file-formats.html#jpeg-saving

Suggested bug fix

In _xobj_to_image(), pypdf could simply provide an extra quality="keep" to Image.save().

IMHO this seems like the best default value,
even if that means a non-fully-backward-compatible change to pypdf.

Feature request

Could it be also possible to introduce a way to provide a custom value for this quality parameter provided to Image.save(), please?

Ideally a new optional argument would be great, but I don't quite see how to make this works with the PageObject.images property that returns a VirtualListImages...

Maybe through a global variable ?
Or through a new parameter of PdfReader?

[stefan6419846]

Thanks for the report. Ideally, we should keep the quality in all cases and let the user provide its own parameters to image.image.save() when having retrieved the image file through PageObject.images.

I currently cannot estimate how this would affect real-world usage, but I consider the current behavior a bug. Thus having a deprecation period should not be required, although we might want to add a corresponding note to the release notes for users which actually rely on the old behavior.

[Lucas-C]

Thank you for your feedback @stefan6419846 👍

I opened PR #3516 to fix this.