Strange pdf with every page referring every image

[HinTak]

Hello. It is the same problem I have been looking at for a while: I have a bunch of pdf's (apparently from Internet Archive) which are generated from scans, and every page is a background image with a mask, at least visually. The mask is the high-def text and the background is just yellowing paper, so extracting the mask as image results in a more readable pdf anyway, and faster/smaller too.

So I have a script which does this:

for pno in range(page_count):
    il = doc.get_page_images(pno)
        for img in il:
              xref = img[1]
                   if (xref > 0):
                         # is mask, do stuff
...

This works okay for every pdf in that collection so far, except 1. That particular one somehow has every page referring to every image. Ie. il is a list of ~600 instead of the 2.

Granted the documentation of doc.get_page_images() says it does not give image per page... so I tried the recommended route, which is page.get_image_info(). Now that's a list of two, but without the actual image data, just dimensions, etc. And page.get_images() returns a list of ~600 like doc.get_page_images(). So what is the api for page.get_image_info() with the actual image data? I tried page.get_textpage().ectractRAWDICT(), that does not seem to help.

[HinTak]

Argh, there is a comment buried in the issue about page.get_images() not returning all images, about my situation of evey page referring to every image: #2714 (comment) - apparently I need to use page.clean_contents(). Let's see if that works...

[HinTak]

Argh, page.clean_contents() does work. I now get a pdf which is about 6MB instead of the 20MB original. Somehow if I join the first and last page of the original with the 2-(n-1) of mod with qpdf, I get a 22MB result. It seems that the first page still refers to every image. Anyway, I think I have got this answered.