Remove link borders and watermark from arXiv papers

clean_arxiv_pdf.py

import pikepdf
import re

def remove_link_borders(pdf):
    '''remove link borders on all pages'''
    removed = 0
    for p, page in enumerate(pdf.pages):
        if '/Annots' in page:
            for a, annot in enumerate(page['/Annots']):
                page['/Annots'][a]['/BS']['/W'] = 0
                removed = removed + 1

    print(str(removed) + ' link borders removed')
    return pdf

def remove_watermark(pdf):
    '''remove the watermark on the first page'''
    '''e.g. arXiv:2108.07732v1  [cs.PL]  16 Aug 2021'''
    '''see https://info.arxiv.org/help/arxiv_identifier_for_services.html'''
    watermark = r'^arXiv:.*\d{4}$'
    page = pdf.pages[0]
    commands = []
    for operands, operator in pikepdf.parse_content_stream(page):
        commands.append([operands, operator])
    
    for c, command in enumerate(commands):
        operands = command[0]
        if operands and isinstance(operands[0], pikepdf.String):
            if re.search(watermark, str(operands[0])):
                del commands[c]
                print('watermark removed')
    new_content_stream = pikepdf.unparse_content_stream(commands)
    page.Contents = pdf.make_stream(new_content_stream)
    return pdf

input_pdf = 'input.pdf'
output_pdf = 'output.pdf'

pdf = pikepdf.Pdf.open(input_pdf)
pdf = remove_link_borders(pdf)
pdf = remove_watermark(pdf)

print('saving to ' + output_pdf)
pdf.save(output_pdf)