Image object not recognized

[flange-ipb]

Description of the bug

I'm extracting images from scientific papers. For this PDF I'm having troubles to extract Fig. 3 on page 10 - this image object is not included in Page.get_images().

I have the same issue in pypdf, see pypdf#3335.

How to reproduce the bug

Run the extract-images scripts on the given PDF file.

PyMuPDF version

1.26.1

Operating system

Linux

Python version

3.13

[JorjMcKie]

This not a bug. There is exactly one image on the page (above "Fig. 2"), which is correctly recognized.
You probably believe that "Fig. 3" also is an image - which is not: it is a vector graphic.

[JorjMcKie]

BTW, what is your underlying problem? Want to extract what looks like an image, no matter what it technically is?

[flange-ipb]

Nice, thanks for the quick analysis.
What I'm designing right now is an agentic AI application that extracts structured data from such publications. One component is to pass images to an optical chemical structure recognition (OCSR) - a task too specific to be handled by multimodal or vision models.
Meanwhile I've also played with pymupdf4llm's to_markdown() function and I'm very happy with the results, in particular it places the image next to the caption which cannot be achieved with Page.get_images() so easily.

[JorjMcKie]

In contrast to many other packages, PyMuPDF can

1. Extract vector graphics
2. Cluster single vectors which are geometrically connected
3. Render parts ("clips") of document pages to images of the desired format

Taking together the above, you can find the rectangle that covers Fig. 3, and make a picture / image of this area.

[JorjMcKie]

Let me invite you to our new Forum, where you can discuss with others (and PyMuPDF's maintainer team) under category "How To".
https://forum.pymupdf.com/c/how-to/6