found multiple images named 'Im0'

[Fake45mar]

Overview

Name of all images extracted on page are the same.
https://pymupdf.readthedocs.io/en/latest/page.html#Page.get_images - item[7] == 'Im0' in all cases
https://pymupdf.readthedocs.io/en/latest/document.html#Document.get_page_images

Repro

image_list = page.get_images()
image_list: [(519, 0, 1162, 1642, 8, 'ICCBased', '', 'Im0', 'DCTDecode'), (544, 0, 583, 1642, 8, 'ICCBased', '', 'Im0', 'DCTDecode'), (595, 594, 144, 80, 8, 'DeviceGray', '', 'Im0', 'DCTDecode')]

Expected behavior

image_list = page.get_images()
image_list: [(519, 0, 1162, 1642, 8, 'ICCBased', '', 'Im0', 'DCTDecode'), (544, 0, 583, 1642, 8, 'ICCBased', '', 'Im1', 'DCTDecode'), (595, 594, 144, 80, 8, 'DeviceGray', '', 'Im2', 'DCTDecode')]

Screenshots

Your configuration (mandatory)

• macOS Monterey 12.5.1
• Python version 3.9.9
• pip3 install wheel
  brew install swig
  brew install freetype
  pip3 install PyMuPDF==1.21.1

[JorjMcKie]

This has nothing to do with (Py-) MuPDF, but is specified that way by the PDF creator.

[Fake45mar]

I can confirm that it's data oriented bug. I ran a batch of documents and not on each got this issue.
Ok, is it any option to obtain image bbox not by the name pdf creator specified?

[JorjMcKie]

If the same image reference name occurs lime "Im0", then it at least should only happen if the image references occur in different XObjects. You can reveal this situation by page.get_images(True). The items of this list are longer and contain the xref of the responsible XObject. If it was the page itself, that last item is 0.

In [1]: import fitz
In [2]: doc=fitz.open("x0.pdf")
In [3]: page=doc[0]
In [4]: page.get_images(True)
Out[4]: [(15, 16, 1200, 1200, 8, 'DeviceRGB', '', 'Im1', 'FlateDecode', 0)]  # image invoked by page itself!