Wrong colors for some images with custom palette

[stefan6419846]

Some images seem to be extracted with the wrong colors. Apparently the custom palette is not considered in these cases. The final byte string from

['/Indexed', '/DeviceRGB', 164, b'\xed\x1c$\xed\x1e&\xed \'\xed (\xed"*\xee#+\xee$,\xee&-\xee\'.\xee(/\xee(0\xee*2\xee,3\xee-4\xee.5\xee/6\xef07\xef29\xef3:\xef4;\xef5<\xef6=\xef7>\xef8?\xef9@\xef:A\xef;B\xf0<C\xf0=D\xf0>E\xf0@F\xf0AH\xf0BI\xf0CJ\xf0EL\xf0FL\xf0HN\xf1IP\xf1JQ\xf1KR\xf1LR\xf1NT\xf1PV\xf1RX\xf1TZ\xf2U[\xf2V\\\xf2X^\xf2Z`\xf2\\b\xf2^d\xf2af\xf3bh\xf3ej\xf3fk\xf3gl\xf3hm\xf3in\xf3jp\xf3lq\xf3ns\xf4ot\xf4pu\xf4rw\xf4sx\xf4uz\xf4v{\xf4w|\xf4x|\xf4y~\xf4z\x7f\xf5{\x80\xf5~\x82\xf5\x80\x84\xf5\x81\x86\xf5\x82\x87\xf5\x84\x88\xf5\x86\x8a\xf6\x88\x8c\xf6\x89\x8e\xf6\x8a\x8e\xf6\x8c\x90\xf6\x8e\x92\xf6\x90\x94\xf6\x92\x95\xf6\x92\x96\xf7\x94\x98\xf7\x97\x9b\xf7\x99\x9c\xf7\x9a\x9d\xf7\x9a\x9e\xf7\x9c\xa0\xf7\x9e\xa1\xf7\x9f\xa2\xf8\xa1\xa4\xf8\xa2\xa5\xf8\xa3\xa7\xf8\xa4\xa7\xf8\xa5\xa8\xf8\xa6\xa9\xf8\xa8\xab\xf8\xa9\xac\xf8\xaa\xad\xf8\xab\xae\xf8\xac\xaf\xf9\xae\xb1\xf9\xb0\xb3\xf9\xb2\xb4\xf9\xb3\xb6\xf9\xb4\xb7\xf9\xb5\xb8\xf9\xb8\xba\xf9\xba\xbc\xfa\xba\xbd\xfa\xbd\xbf\xfa\xbe\xc0\xfa\xc0\xc2\xfa\xc2\xc5\xfa\xc4\xc6\xfb\xc7\xc9\xfb\xc8\xca\xfb\xca\xcb\xfb\xca\xcc\xfb\xcc\xce\xfb\xce\xd0\xfb\xd0\xd1\xfb\xd1\xd2\xfb\xd2\xd3\xfb\xd2\xd4\xfc\xd3\xd5\xfc\xd4\xd6\xfc\xd6\xd7\xfc\xd7\xd8\xfc\xd8\xd9\xfc\xd9\xda\xfc\xdb\xdd\xfc\xdc\xde\xfc\xde\xdf\xfc\xdf\xe0\xfd\xe0\xe1\xfd\xe1\xe2\xfd\xe3\xe4\xfd\xe4\xe5\xfd\xe5\xe6\xfd\xe6\xe7\xfd\xe7\xe8\xfd\xe8\xe9\xfd\xe9\xea\xfd\xea\xea\xfd\xeb\xec\xfe\xec\xed\xfe\xed\xee\xfe\xee\xef\xfe\xef\xf0\xfe\xf1\xf1\xfe\xf3\xf3\xfe\xf3\xf4\xfe\xf4\xf5\xfe\xf5\xf6\xfe\xf6\xf6\xfe\xf8\xf8\xff\xfa\xfa\xff\xfb\xfc\xff\xfc\xfc\xff\xfe\xfe']

apparently gets ignored, as img.putpalette(color_space[-1]) will yield correct colors (red instead of the currenty gray returned by pypdf).

Note: This has been originally reported as (2) in https://github.com/py-pdf/pypdf/issues/2303#issuecomment-1823199023, but due to being a separate issue, I just isolated it.

Environment

Which environment were you using when you encountered the problem?

$ python -m platform
Linux-5.14.21-150400.24.97-default-x86_64-with-glibc2.31

$ python -c "import pypdf;print(pypdf._debug_versions)"
pypdf==3.17.1, crypt_provider=('pycryptodome', '3.18.0'), PIL=10.0.0

Code + PDF

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

from pypdf import PdfReader

for index, page in enumerate(PdfReader('out1.pdf').pages):
    print(index, page)
    for key in page.images.keys():
        print(key)
        print(page.images[key].indirect_reference)
        page.images[key].image.convert("RGB").save("image.png")

The PDF file is the same as in https://github.com/py-pdf/pypdf/issues/2303#issue-2002452594, although it requires some additional work due to the EOD issue there.

[pubpub-zz]

@stefan6419846 I can't (anymore ?) reproduce this issue : the image is red colored . Can you confirm and close it if so ?

[stefan6419846]

Yes, this seems to have indeed been fixed in the meantime. The top-down representation is unrelated to this case.

[pubpub-zz]

The top-down representation is unrelated to this case.

There is no issue about the top-down : the image is like this and is just flipped in the page

[stefan6419846]

Yes, no worries - this has not been part of the ticket anyway ;)

[pubpub-zz]

however closing this issue lower the number of open issues below 75!🎉🎉🎉🎉