Strange coordinates and size of newly inserted images

[said46]

In some pdf document the coordinates of a new inserted image via Page.insert_image() function are very strange.
Code:

from fitz import Page, Document

doc: Document = Document("example_pdfs/bad stamp example.pdf")
page: Page = doc[0]
print(f"{page.rect=}")

page.insert_image((0, 0, 400, 200), filename='images/RLMU_Stamp.png', keep_proportion=True,
                  overlay=True, rotate=page.rotation)

images = page.get_images()

page_image_xref = images[0][0]
stamp_image_xref = images[1][0]

new_stamp_bbox = page.get_image_rects(images[1][7])
print(f"{new_stamp_bbox=}")

doc.save('example_pdfs/test.pdf')
doc.close()

Output:

page.rect=Rect(0.0, 0.0, 1189.0, 841.0)
new_stamp_bbox=[Rect(0.0, -696915.375, 475600.0, -547764.5)]

LINK TO THE PDF: pdf link

EDIT: I could not find a way to move the already inserted image via PyMuPDF (hundreds of pdfs), found no examples or explanation.

[JorjMcKie]

This question has been answered in the associated issue #2495.

[JorjMcKie]

Test 1:

In [1]: import fitz
In [2]: doc=fitz.open("bad stamp example.pdf")
In [3]: page=doc[0]
In [4]: page.wrap_contents()
In [5]: xref = page.insert_image((0, 0, 400, 200), filename="nur-ruhig.png")
In [6]: # save PDF to memory and re-open:
In [7]: pdfdata = doc.tobytes(deflate=True)
In [8]: doc.close()
In [9]: doc = fitz.open("pdf", pdfdata)
In [10]: page = doc[0]
In [11]: page.get_image_rects(xref)
Out[11]: [Rect(112.375244140625, 0.0, 287.624755859375, 200.0)]
In [12]: # PERFECT!!!

[JorjMcKie]

I suggest you use the following script. It should able to cope with all or most peculiar input files (rotated pages, weird geometries, weird MediaBox / CropBox and what not).

import fitz

doc = fitz.open()  # new output file
src = fitz.open("2.pdf")  # source file
imgfile = "nur-ruhig.png"
for src_page in src:  # iterate over input pages
    src_rect = src_page.rect  # source page rect
    w, h = src_rect.br  # save its width, height
    src_rot = src_page.rotation  # save source rotation
    src_page.set_rotation(0)  # set rotation to 0 temporarily
    page = doc.new_page(width=w, height=h)  # make output page
    page.show_pdf_page(  # insert source page
        page.rect,
        src,
        src_page.number,
        rotate=-src_rot,  # use reversed original rotation
    )
    page.insert_image((0, 0, 400, 200), filename=imgfile)

doc.ez_save("output.pdf")

Every output page will have rotation 0 and all images shold be inserted as expected.

[JorjMcKie]

Things like page.wrap_contents() and similar are not required when you use this alternative, because it internally does that: the output page shows the input page wrapped in "q"/"Q" commands.