Pixmap.save Unable to save pNG format correctly

[jianghwjx]

I used python to write a pdf format invoice into a png image program, found that the new version of PyMuPDF Pixmap.save () save the png format may appear to stamp opaque problem, try to use 1.20.0 before the version of the library, use Pixmap.writePNG () save is normal, I would like to ask about the new I would like to ask the new version there are other methods equivalent to Pixmap.writePNG command to achieve the same function?

[JorjMcKie]

You are referring to method names that have been changed a long time ago now.
Saving a pixmap to a file now simply is pix.save(filename). Please note that JPEG is now among the image formats that are supported by PyMuPDF natively. So if the filename endswith ".jpg" or ".jpeg" JPEG format is used.

Saving to memory is now called pix.tobytes("format"), where format is any of the supported output format as above.

[jianghwjx]

Using the original command to save PNG normally should look like this

[JorjMcKie]

The image with the red stamp obviously is not a transparent image obviously. So when it was created from a Pixmap that Pixmap did not have an alpha channel, pix.alpha = 0.
Why this was or was not the case depends on your upstream app creating the pixmap.
Otherwise, only PNG images can be transparent - JPGs never are.

[jianghwjx]

The PDF file uses this
通行费.pdf
The same PDF uses the following code, version 1.18.19, and saving the PNG is transparent

def to_png(input_file):
    name = input_file[:-4]
    name2 = input_file.split('\\')[-1]
    pdf = fitz.open(input_file)
    rname = ''

    for page_num in tqdm(range(pdf.page_count), desc=f'正在转换：[{name2}]'):
        pix = pdf[page_num].get_pixmap(matrix=fitz.Matrix(3, 3))
        rname = f"{name}_page_%i.png" % (page_num + 1)
        pix.writePNG(rname)

The new version is opaque when saved using pix.save

def to_png(input_file):
    name = input_file[:-4]
    name2 = input_file.split('\\')[-1]
    pdf = fitz.open(input_file)
    rname = ''

    for page_num in tqdm(range(pdf.page_count), desc=f'正在转换：[{name2}]'):
        pix = pdf[page_num].get_pixmap(matrix=fitz.Matrix(3, 3))
        rname = f"{name}_page_%i.png" % (page_num + 1)
        # pix.writePNG(rname)
        pix.save(rname)

Because our English is not very good, is it a problem with my understanding, or is it that pix.save does not support saving PNG format

[JorjMcKie]

is it a problem with my understanding,

I think so, yes. Whether or not a pixmap has an alpha channel is defined at creation time. Please look at the documentation for Page.get_pixmap(). When you say alpha=True the pixmap will have an alpha channel. The default is alpha=False.
I don't remember whether the default under version 1.18.x was any different.

But saving a pixmap will not change transparency.

[jianghwjx]

First of all, thank you for your patient reply
Actually, I have already tried alpha=True before giving feedback on this issue.
Below is the png for saving the execution results of version 1.18.19

    for page_num in tqdm(range(pdf.page_count), desc=f'正在转换：[{name2}]'):
        pix = pdf[page_num].get_pixmap(matrix=fitz.Matrix(3, 3), alpha=True)
        rname = f"{name}_page_%i.png" % (page_num + 1)
        pix.writePNG(rname)
        # pix.save(rname)

The saved image looks like this

However, using the new version, the code is as follows

    for page_num in tqdm(range(pdf.page_count), desc=f'正在转换：[{name2}]'):
        pix = pdf[page_num].get_pixmap(matrix=fitz.Matrix(3, 3), alpha=True)
        rname = f"{name}_page_%i.png" % (page_num + 1)
        # pix.writePNG(rname)
        pix.save(rname)

The saved image looks like this

The above tests used the same PDF file
通行费.pdf

[JorjMcKie]

Cannot reproduce. The following snippet works exactly as it should:

doc=fitz.open("default.pdf")
page=doc[0]
pix=page.get_pixmap(matrix=fitz.Matrix(3,3),alpha=True)
pix.save("test.png")
fitz.__version__
'1.23.7'

With alpha=False:

[jianghwjx]

The situation is indeed a bit strange, this is the result of my test

and

Not sure what other information I can provide to locate the problem

[JorjMcKie]

Well best would be everything that you haven't mentioned yet, notably operating system and PyMuPDF installation method.

Found the problem:
It is connected with an issue in the base library MuPDF. My local PyMuPDF version already uses the current, corrected MuPDF, therefore I am able to produce correct images.

So you will have to wait with your upgrade until PyMuPDF v1.23.8 comes out.

[jianghwjx]

Thank you very much for your answer, you are really very patient, thank you again

[jianghwjx]

Has version 1.23.8 not fixed this issue yet
After attempting to upgrade, this issue still exists

[jianghwjx]

这个问题是不是还没有在新版本中修正，因为我尝试使用1.23.8及以后的版本，依然存在这个问题，能确认一下吗
Has this issue not been fixed in the new version yet? I tried using versions 1.23.8 and later, but the issue still exists. Can you confirm

[JorjMcKie]

Yes, unfortunately.
The fix must happen in MuPDF, not in PyMuPDF. The MuPDF team is still working on this.

[jianghwjx]

Okay, then I'll keep paying attention. Thank you.

[yash-0501]

Is this issue, fixed yet? @JorjMcKie, I am facing the same kind of issue