如何用Python结合pdfminer提取PDF文档的页眉和页脚？

Hey there! Since you're already using pdfminer to pull text from your PDF, detecting headers and footers is totally feasible with a few targeted, practical approaches. Let me walk you through the most reliable ones:

Headers and footers almost always live in the extreme top or bottom regions of a page. The key here is that pdfminer lets you access the coordinates (x0, y0, x1, y1) of every text block it extracts.

Here's how to implement this:

• First, get the total height of the page using the PDF's mediabox (this defines the page's boundaries).
• Define threshold ranges for header and footer areas—for example, the top 10% and bottom 10% of the page height.
• Iterate through all text blocks on each page, and check if their vertical position falls within these ranges.

A quick code snippet to illustrate:

from pdfminer.layout import LAParams, LTTextBoxHorizontal
from pdfminer.pdfpage import PDFPage
from pdfminer.pdfinterp import PDFResourceManager, PDFPageInterpreter
from pdfminer.converter import PDFPageAggregator

def detect_header_footer_by_location(page_layout, page_height, header_percent=0.1, footer_percent=0.1):
    header_threshold = page_height * header_percent
    footer_threshold = page_height * footer_percent
    
    header_text = []
    footer_text = []
    
    for element in page_layout:
        if isinstance(element, LTTextBoxHorizontal):
            # y1 is the top coordinate of the text block
            if element.y1 >= page_height - header_threshold:
                header_text.append(element.get_text().strip())
            # y0 is the bottom coordinate of the text block
            elif element.y0 <= footer_threshold:
                footer_text.append(element.get_text().strip())
    
    return '\n'.join(header_text), '\n'.join(footer_text)

# Example usage setup
rm = PDFResourceManager()
laparams = LAParams()
device = PDFPageAggregator(rm, laparams=laparams)
interpreter = PDFPageInterpreter(rm, device)

with open("your_document.pdf", "rb") as fp:
    for page in PDFPage.get_pages(fp):
        interpreter.process_page(page)
        layout = device.get_result()
        page_height = page.mediabox[3] - page.mediabox[1]
        header, footer = detect_header_footer_by_location(layout, page_height)
        print(f"Header: {header}\nFooter: {footer}")

Headers and footers typically repeat across most (if not all) pages—think page numbers, document titles, or standardized branding text.

Here's the workflow:

• For each page, extract candidate text from top/bottom regions using the location method above.
• Track how often each text segment appears across pages.
• Any segment that appears on a large majority of pages (e.g., 80%+ is a good rule of thumb) is almost certainly a header or footer.
• Note: For dynamic content like page numbers, normalize the text (e.g., remove digits or replace them with a placeholder) before checking repetition.

Some PDFs are tagged (Tagged PDF) with semantic structure labels—this includes explicit tags for headers, footers, body text, etc. Pdfminer can parse these structural elements directly if the PDF supports it.

You'll need to use pdfminer's PDFStructTreeParser to traverse the document's structure tree and look for nodes labeled as "Header" or "Footer". Keep in mind this only works for properly tagged PDFs, which aren't universal.

• Odd/Even Page Differences: Some documents have different headers/footers on odd vs even pages (e.g., page numbers on left vs right). Split your detection logic to handle odd and even pages separately.
• Noise Reduction: Ignore small, one-off text snippets in header/footer regions (like a single stray word) by filtering out text below a certain length.
• Combine Methods: For the most accuracy, use both location and repetition checks—e.g., first narrow down candidates by position, then verify they repeat across pages.

内容的提问来源于stack exchange，提问作者Anil Yadav