Changing background color of existing PDF page

[usmansc]

I have a pdf from external source. Sometimes it has background color, sometimes it does not. I want it to be without the background color. I am trying to do this manually for ages, today I found pymupdf and I am wondering if such a thing is possible with it. I am not sure how it should be done with pymu, but manually I was trying to find the object that represented the background, delete the object and update xref table. Are such things possible with pymu?

[JorjMcKie]

I have a pdf from external source. Sometimes it has background color, sometimes it does not. I want it to be without the background color. I am trying to do this manually for ages, today I found pymupdf and I am wondering if such a thing is possible with it. I am not sure how it should be done with pymu, but manually I was trying to find the object that represented the background, delete the object and update xref table. Are such things possible with pymu?

Yes, this is possible.
As always with PDF, there are amazingly many ways to achieve the type of thing you describe (incomplete list):

• Image used covering the page. Solution: find the image xref and replace with a small, empty, transparent image.
• PDF "line art" knows ways to generate filled rectangles. This may be very hard to detect - sometimes even impossible with any common sense effort. If these instructions can be detected however, then PyMuPDF has ways to delete them.
• PDF Watermark Artifacts are also part of the PDF spec. Easier to detect and remove than the previous case.

As you mentioned xref, your cases seem to belong to the first point above, though.
Please add more detail to your question.

[JorjMcKie]

Forgot to mention:
To also physically get rid of the green image, use a save option that recovers its space: doc.ez_save() or doc.save(filename, garbage=3,...).

[JorjMcKie]

It actually has no watermark, but drawings that look like text - but are not (that diagonal transparent name).
To delete this type of thing is practicaly impossible.

[nissansz]

How can I draw an image or white rectangular on a specified part of page?

[JorjMcKie]

Please do read the documentation!

• page.draw_rect() draws rectangle, there exist many more similar to draw other objects.
• page.insert_image() insert an image
• etc.

[nuthinking]

@JorjMcKie once all images are removed, is it possible to export a transparent PNG? I tried with page.get_pixmap(matrix=mat, alpha=True) and I still get the white background. I removed all xrefs and images are not visible anymore. I would like to export all text with a transparent background, basically. Thanks!

[JorjMcKie]

@JorjMcKie once all images are removed, is it possible to export a transparent PNG? I tried with page.get_pixmap(matrix=mat, alpha=True) and I still get the white background. I removed all xrefs and images are not visible anymore. I would like to export all text with a transparent background, basically. Thanks!

You can make such an unremovable background yourself, e.g. page.draw_rect(page.rect, color=None, fill=fitz.pdfcolor["yellow"], overlay=False).
If you are in a hacky mood, you may want to look at the page's /Contents object(s). At its start, some instruction doing this should be present ... which you can remove by editing and updating that object ...

[JorjMcKie]

You cannot simply "edit" any PDF directly. That destroys the inner structure in a catastrophic way. Instead you must extract what you want to change, edit this and then re-insert (update). This leaves MuPDF a change to get all those pointers updated.

[nuthinking]

It corrupts even when I edit->undo->save. Meaning not making any changes. So I thought was a problem with file encoding. What a nasty format! :D

[JorjMcKie]

Absolutely yes. You probably also have a nasty editor (Mac!).

[nuthinking]

What would be an "empty" /Contents content I could replace mine with?

[nuthinking]

Using TextEdit doesn't break it. Was Sublime.

[JorjMcKie]

What would be an "empty" /Contents content I could replace mine with?

If xref is the xref number of a /Contents object, you can empty it by simply doing doc.update_stream(xref, b"").

[nuthinking]

Looks like the MediaBox is useful. No idea what else to remove from here:

<</Type/Page/Resources 11 0 R/MediaBox[0 7.8299815 595.5 850.07998]/Contents[12 0 R 179 0 R 180 0 R 181 0 R 182 0 R 183 0 R 184 0 R 185 0 R 186 0 R 187 0 R]/StructParents 0/Parent 2 0 R/Tabs/S/Rotate 0/Annots[]>>

I already removed the various BleedBox, TrimBox and CropBox 🤷‍♂️

[JorjMcKie]

You mean in order to remove that background still?
Those poor boxes have nothing to do with it!
The code that causes the white background is very probably contained in xref 12 (see "12 0 R" above).

[JorjMcKie]

Why don't let me have that PDF so I can take a look a it ...

[nuthinking]

Check your email 🙏