Wrong rect of Document when open "bmp" of "png" files by fitz.open()

[Jianpeng-teng]

Please provide all mandatory information!

Describe the bug (mandatory)

Open bmp or png files (maybe also other kind of pictures) by fitz.open(file_adress) will return a Document with wrong rect. Which is small than what it should be with a ratio of 0.75.

To Reproduce (mandatory)

import fitz
img1 = fitz.open("E:\pic\0.png") # 0.png is (1024 x 768)
rect = img1[0].rect # rect is ( 0, 0, 768, 576)

you may see the x_zoom == y_zoom == (768/1024) == (576/768) == 0.75

And it does not relate to the size of the picture. No matter what picture, the zoom is always 0.75

Your configuration (mandatory)

• Operating system, potentially version and bitness

Window10 (19044.2130)

• Python version, bitness

Python 3.8.5

• PyMuPDF version, installation method (wheel or generated from source).

1.20.2 (I tried 1.20.0 and 1.20.1 also. same)
installed with

For example, the output of print(sys.version, "\n", sys.platform, "\n", fitz.__doc__) would be sufficient (for the first two bullets).

Additional context (optional)

Add any other context about the problem here.

[JorjMcKie]

This is no bug, but a misconception about how page sizes of Images are computed when opened as a document.

[JorjMcKie]

I will transfer the issue to a Discussion and answer it there.

[JorjMcKie]

Thanks for asking ... this may be confusing, but has a simple explanation.

For example take this image:
Its dimension is 607 x 741 pixels, its DPI is 120 x 120.

When opened as a document, the corresponding page size is:

page.rect
Rect(0.0, 0.0, 364.20001220703125, 444.6000061035156)

The computation works as follows, taking rect.width as an example:

(img width in pixels) / (x-dpi) * (points per inch)  = page.rect.width
      607             /   120   *        72          = 364.2

Or conversely:

page.rect.width * 120 / 72
607.0000203450521
# the original image x-pixels

[Jianpeng-teng]

Wait, I suddenly realized that, so the Rect size will decrease as DPI increases? That's confusing.

[Jianpeng-teng]

Sorry, I was not in front of the screen. that is helpful. So actually it’s happened because the default DPI has been set to 96x96? It is kind of tricky cause when I try to transfer the pdf document to png and do manipulation and transfer them back, I realized the whole document become smaller than before. 获取 Outlook for iOS<https://aka.ms/o0ukef>

…

________________________________ 发件人: Jorj X. McKie ***@***.***> 发送时间: Sunday, October 16, 2022 5:30:22 PM 收件人: pymupdf/PyMuPDF ***@***.***> 抄送: Jianpeng-teng ***@***.***>; Author ***@***.***> 主题: Re: [pymupdf/PyMuPDF] Wrong rect of Document when open "bmp" of "png" files by fitz.open() (Discussion #1969) Thanks for asking ... this may be confusing, but has a simple explanation. For example take this image: Its dimension is 607 x 741 pixels, its DPI is 120 x 120. [grafik]<https://user-images.githubusercontent.com/8290722/196023365-265ef192-955a-44b8-838c-ef7dcc688b2f.png> When opened as a document, the corresponding page size is: page.rect Rect(0.0, 0.0, 364.20001220703125, 444.6000061035156) The computation works as follows, taking rect.width as an example: (img width in pixels) / (x-dpi) * (points per inch) = page.rect.width 607 / 120 * 72 = 364.2 Or conversely: page.rect.width * 120 / 72 607.0000203450521 # the original image x-pixels ― Reply to this email directly, view it on GitHub<#1969 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AOLKXQKDA2YEP65USMY6XNLWDOVI5ANCNFSM6AAAAAARGHTEWM>. You are receiving this because you authored the thread.Message ID: ***@***.***>

[JorjMcKie]

when I try to transfer the pdf document to png

A little unclear what you are trying to do. But you can influence the PNG size of the PDF page's image (the Pixmap) by choosing an appropriate DPI value. You will of course have that problem, if you leave everything to default values.

[JorjMcKie]

When making the page pixmap use the dpi parameter: pix = page.get_pixmap(dpi=nnn). Choose the value nnn according to the above formula, so the PNG opened as document again has the original page dimensions.

[Jianpeng-teng]

Thank you so much!
Actually, I just want to crop the pdf page to a new size, like, the margin of some of the pages were too wide. I found the idea on this page:
https://pymupdf.readthedocs.io/en/latest/recipes-images.html#how-to-create-partial-pixmaps-clips

trans = fitz.Matrix(zoom_x, zoom_y)
clip = fitz.Rect(x1,y1,x2,y2)
pix = page.get_pixmap(matrix=trans, clip=clip)

And then I couldn't find a simple way to make the pixmap directly back to pdf page. So:

pix.save(page_number + ".png")
doc = fitz.open()
imgdoc = fitz.open(page_number + ".png")
pdfbytes = imgdoc.convert_to_pdf()
imgdoc.close()
imgpdf = fitz.open("pdf", pdfbytes)
doc.insert_pdf(imgpdf)

(By the way, neither insert_page() nor show_pdf_page() works said in this tutorial:
https://pymupdf.readthedocs.io/en/latest/recipes-images.html#how-to-add-images-to-a-pdf-page
I realized that they had been marked as "private":

Also, the same thing happened to new_page() in:
https://pymupdf.readthedocs.io/en/latest/recipes-general.html#how-to-add-pages

Are they all replaced by:

insert_pdf()

?

[JorjMcKie]

Okay, so all you want is making the visible part of some page smaller?

If this is true, then using pixmaps and images simply is the wrong idea!
Because a page Image is something very different from the original page - for example example, all the text is gone, and zooming into the page will show single pixels at some point and more of that kind.

All you need to do is adjusting its CropBox via page.set_cropbox().

[Jianpeng-teng]

Ohhh nooo. What am I doing....