Extract an entire image, without splitting it into smaller ones

[molokanov50]

Hello,

To extract images from a PDF, I use the code according to your readme (see Recipes -> Extract all images from a PDF). But instead the entire upper image (see page 4 of the attached file), I receive 15 smaller ones, with various icons and black rectangles being treated as separate images.
In the other attachment, I showed the entire image I would like to extract. Note that the image area has some text layer bboxes. I partially highlighted the text layer to make the situation clear.
Is it possible in PyMuPDF to extract this image entirely and not to add smaller images included in the entire image area to the result? If yes, what code / which options should I use? If not, why?

Externalization.pdf

[JorjMcKie]

As the displayed text on the page suggests: there is no such thing as the "entire image" - it is a cognitive artifact created by the mind of the viewer - i.e. by you.
Technically, there exists a mixture of text, vector graphics and images in that region. And even viewers may have different opinions about what we are looking at: maybe two "images" - each of the yellow regions being a separate one?
These are all separate pdf objects. There exists no information whatsoever that they semantically belong together.
If you extract the page with pymupdf4llm you will probably (haven't confirmed it yet) find that its layout function has classified the area as a "figure" and saved it like a screen print using the Pixmap of the area.

[JorjMcKie]

Moving this Discussions because no library issue is present.

[Samuel-Rodriguezs]

Adding a practical workaround to Jorj's explanation: since the "entire image" is really a mix of several images, vector graphics and text, you can render that region of the page instead of extracting the embedded objects. You get exactly what you see:

import pymupdf

doc = pymupdf.open("Externalization.pdf")
page = doc[3]  # page 4

# bounding boxes of all images drawn on the page
rects = [pymupdf.Rect(info["bbox"]) for info in page.get_image_info()]

# keep only the pieces of the area you want (here: upper half of the page)
pieces = [r for r in rects if r.y1 < page.rect.height / 2]

area = pymupdf.Rect(pieces[0])
for r in pieces[1:]:
    area |= r  # union of all pieces

pix = page.get_pixmap(clip=area, dpi=200)  # raise dpi for more detail
pix.save("entire-image.png")

The result is a single PNG of that region, including the icons, rectangles and the text drawn over it. If you already know the coordinates, you can skip the image boxes and pass clip=pymupdf.Rect(x0, y0, x1, y1) directly.