How to remove the contents of the images?

[yusheng0104]

Hi,

Is there a method could remove all the contents of an image in a pdf file? Pymupdf is sometimes too powerful and picks up some words from images. And my goal is analyzing the text not in the image.

Thanks

[JorjMcKie]

I think I don't understand your problem.
There are several ways how to ignore images and only extract text: page method get_text has an option string as argument one, and a keyword parameter flags, which together control this:

• option = "text" will only extract text
• switching off TEXT_PRESERVE_IMAGES from flags will only extract text for the other option values ("blocjs", "dict" etc.)

Please be more specific.

[yusheng0104]

Thanks. You've answered my question. I was bothered by some weird text from the images.

[JorjMcKie]

Ok, good! That was easy then 😎

[yusheng0104]

Could also exclude the metadata of the images? In the documentation, it says TEXT_PRESERVE_IMAGES will return metadata.
Thanks,

[JorjMcKie]

If bit TEXT_PRESERVE_IMAGES is switched off, no image information at all will be included in any output type. Also no meta information.

[yusheng0104]

Thanks a lot. Could you please teach me how to switch TEXT_PRESERVE_IMAGES off?

[JorjMcKie]

Use the Python "OR" operator "|" to set single bits of the flags argument.
Depending on the text method used, there are default combinations used, if you do not specify flags - see here.
Otherwise use this list to determine the bits you need:

flags = (fitz.TEXT_PRESERVE_LIGATURES |
    fitz.TEXT_PRESERVE_WHITESPACE
    )

Then use flags in your text extraction or search function. The above example will ignore all images.

[yusheng0104]

Thanks. I tried get_text(options = "text", flags =3), but still got text from the images.

[JorjMcKie]

I am sure there is some misconception: this extraction format does never access image information!
Can you give me the file?
If you have confidentiality concerns, you may use my private e-mail [email protected].