Detect a "DRAFT" watermark?

[mnoah66]

Is it possible to just detect if "DRAFT" watermark exists? I have seen some google search results that want to remove or replace it. I want neither.

Just detect :)

I've tried drawing rectangle and getting text:

draft = fitz.Rect(65, 120, 520,  745)
print("DRAFT" in page.get_textbox(draft)) # False

PDF image. The pdf has medical info and cannot be shared

[JorjMcKie]

Hard to give a definitive answer if you cannot share an example file. But maybe your watermark can be removed by the recipe I developed here.

[JorjMcKie]

Interesting.
Then maybe the image is inserted on the page "traditionally" - i.e. not using PDF watermark technology.
Depending on how you extracted the image, xref 3 may contain it. Check this by print(page.get_images()).
If so, then you can simply replace the image by a small invisible other image using this example script.

You have no chance to share an example file? Even to my personal e-mail address?

[mnoah66]

I emailed you the file. I am not looking to replace the image/watermark. Just detect if it is there.

Thank you! :)

print(page.get_images())
[(3, 0, 816, 912, 8, 'DeviceRGB', '', 'Im3', 'FlateDecode'), (5, 0, 33, 83, 8, 'DeviceRGB', '', 'Im5', 'FlateDecode'), (7, 0, 130, 128, 8, 'DeviceRGB', '', 'Im7', 'FlateDecode'), (9, 0, 1329, 106, 8, 'Indexed', '', 'Im9', 'FlateDecode'), (10, 0, 916, 463, 8, 'DeviceRGB', '', 'Im10', 'FlateDecode')]

```

[JorjMcKie]

Thanks for the file.
As suspected, this is no watermark - just a transparent image. As such, you cannot definitely tell what its role is - you know nothing about it except the few metadata contained in dictionary doc.extract_image(3).
You can also determine, where in which location on which page this image is being shown.
That's about it. Strictly speaking, this is no watermark.

[mnoah66]

@JorjMcKie thanks again. I think I have enough to work with given that it is the only image with a height of over 900. I could loop over the images and check for the height and deduce if it is a draft. I think adding OCR would be overkill.

[JorjMcKie]

I think adding OCR would be overkill.

I tried that already. Neither Tesseract, nor OCRmyPDF (which is based on that anyway) can decipher that image.