pdf clean script

clean_pdf.py

import fitz  # PyMuPDF
import os
from PIL import Image
from tqdm import tqdm
import cv2
import numpy as np
import img2pdf

H = 1280
def pdf_to_png(*args):
    imgs = []
    pdf_path, output_folder = args
    # 打开 PDF 文件
    pdf_document = fitz.open(pdf_path)

    # 遍历每一页，保存为 PNG 图片
    pbar = tqdm(range(pdf_document.page_count), desc=f"Converting {pdf_path.split('/')[-1]}")
    for page_number in pbar:
        page = pdf_document[page_number]
        image = page.get_pixmap()
        h = image.height
        ratio = H / h
        # we need high enough resolution
        if ratio > 1:
            image = page.get_pixmap(matrix=fitz.Matrix(ratio, ratio))
        
        # 转换为 Pillow 图像
        pil_image = Image.frombytes("RGB", [image.width, image.height], image.samples)
        np_image = np.array(pil_image.convert('L'))
        # _, processed = cv2.threshold(np_image, 150, 255, cv2.THRESH_BINARY)
        np_image[np_image > 200] = 255
        np_image[np_image < 127] = 0
        # np_image[(np_image >= 127) & (np_image <= 200)] = 127
        imgs.append(np_image)

        # _,processed = cv2.threshold(np_image,0,255,cv2.THRESH_BINARY+cv2.THRESH_OTSU)
        
        # 保存为 PNG 图片
        output_path = f"{output_folder}/{page_number + 1}.png"
        cv2.imwrite(output_path, np_image)

    # 关闭 PDF 文件
    pdf_document.close()
    image_folder = output_folder
    image_files = [f for f in os.listdir(image_folder) if f.endswith('.png')]
    image_files.sort(key=lambda x: int(x.removesuffix('.png')))  # 确保按顺序处理图像

# 将所有图像转换为PDF
    pdf_bytes = img2pdf.convert([os.path.join(image_folder, f) for f in image_files])
    with open('./output2.pdf', 'wb') as f:
        f.write(pdf_bytes)

if __name__=='__main__':
    pdf_to_png('./音韵学概论_麦耘.pdf', './output/')