Some scan pdfs' text is wrong, i think the text is gotten from ocr, so how can l get the original background image?

[njustczr]

[JorjMcKie]

This has no description at all.
The title suggests you may have wanted to ask a question in "Discussions".
If so, please open a post there - and with sufficient background please.

[njustczr]

This has no description at all. The title suggests you may have wanted to ask a question in "Discussions". If so, please open a post there - and with sufficient background please.

the pdf is

but i copy the text is "第丸章 肘足虫"
the text is gotten by OCR, so if i can get the scan background image

[JorjMcKie]

You can either use a package like OCRmyPDF to force-rescan the document, or use PyMuPDF's built-in OCR feature.
In both cases you should make sure to specify the presence of language Chinese.

You can also look at the page's images page.get_images() and extract them via doc.extract_image(). But I recommend trying to re-OCR the page instead. Use a large dpi to improve the recognition quality.

[njustczr]

I have tried the following code to save the background image,but i can only get a white image.
# doc = fitz.open(pdf_path)
# page = doc[10]
# xrefs = page.get_images()
# print(xrefs)
# idx = 0
# for xref in xrefs:
# idx += 1
# pix = fitz.Pixmap(doc, xref[0])
# if pix.n < 5:
# pix.save(str(idx) + '.png')
# else:
# pix1 = fitz.Pixmap(fitz.csRGB, pix)
# pix1.save(str(idx) + ".png")
# pix1 = None
# pix = None

[njustczr]

You can either use a package like OCRmyPDF to force-rescan the document, or use PyMuPDF's built-in OCR feature. In both cases you should make sure to specify the presence of language Chinese.

You can also look at the page's images page.get_images() and extract them via doc.extract_image(). But I recommend trying to re-OCR the page instead. Use a large dpi to improve the recognition quality.

thanks