使用fitz的clip参数提取PDF文本返回空值，求解决方法

我正在编写Python脚本将PDF转换为有声书，尝试通过设置边框（clip参数）移除页码及其他无关标题。当前使用Python 3.10.4，使用无参数的get_text可正常提取文本，但传入clip参数后返回空值。我使用的输入参数为起始页27、结束页40、边框比例0.05，代码如下：

import fitz
from gtts import gTTS
import os
import re

def extract_text_by_area(page, x0, y0, x1, y1):
    return page.get_text("text", clip=(x0, y0, x1, y1))

def get_text_with_area_extraction(filepath: str, start_page: int, end_page: int, x0, y0, x1, y1) -> str:
    with fitz.open(filepath) as doc:
        extracted_text = ""
        for page_num in range(start_page, end_page + 1):
            page = doc[page_num]
            extracted_text += extract_text_by_area(page, x0, y0, x1, y1).replace('\n', ' ').strip()
        
        standardized_text = re.sub(r'\s+', ' ', extracted_text)
        return standardized_text
    
def get_pdf_dimensions(pdf_path, page_num):
    doc = fitz.open(pdf_path)
    
    page = doc[page_num]
    page_width = page.rect.width
    page_height = page.rect.height
    
    doc.close()
    return page_width, page_height

def calculate_text_area(page_width, page_height, percentage):
    border_x = page_width * percentage
    border_y = page_height * percentage

    x0 = border_x
    y0 = page_height - border_y
    x1 = page_width - border_x
    y1 = border_y

    
    return x0, y0, x1, y1

filepath = 'Fooled-by-Randomness-Role-of-Chance-in-Markets-and-Life-PROPER1.pdf'
start_page = int(input("Enter the starting page: ")) - 1
end_page = int(input("Enter the ending page: ")) - 1

page_num = start_page
page_width, page_height = get_pdf_dimensions(filepath, page_num)

percentage = float(input("Enter the percentage of the border to remove (e.g., 0.1 for 10%): "))
x0, y0, x1, y1 = calculate_text_area(page_width, page_height, percentage)
	extracted_text = get_text_with_area_extraction(filepath, start_page, end_page, x0, y0, x1, y1)

print(extracted_text)

tts = gTTS(text=extracted_text, lang='en')

output_audio_path = 'extracted_text_audio.mp3'
tts.save(output_audio_path)

print(f"Text-to-speech audio saved as: {output_audio_path}")

核心问题：Clip坐标顺序错误 #

PyMuPDF中rect的坐标规则是**(x0, y0, x1, y1)**，其中：

• x0: 左边界，x1: 右边界（x1 > x0）
• y0: 上边界，y1: 下边界（y1 > y0）

你的calculate_text_area函数把y0和y1搞反了，导致clip区域是倒置的无效区域，因此提取不到文本。修改该函数：

def calculate_text_area(page_width, page_height, percentage):
    border_x = page_width * percentage
    border_y = page_height * percentage

    x0 = border_x
    y0 = border_y  # 上边界：从顶部往下留border_y距离
    x1 = page_width - border_x
    y1 = page_height - border_y  # 下边界：从底部往上留border_y距离
    
    return x0, y0, x1, y1

其他优化建议 #

• 适配多页面尺寸差异：部分PDF的不同页面可能有不同尺寸，建议在提取每页文本时单独计算当前页的clip区域，而非仅使用起始页尺寸：

  def get_text_with_area_extraction(filepath: str, start_page: int, end_page: int, percentage) -> str:
      with fitz.open(filepath) as doc:
          extracted_text = ""
          for page_num in range(start_page, end_page + 1):
              page = doc[page_num]
              page_width = page.rect.width
              page_height = page.rect.height
              x0, y0, x1, y1 = calculate_text_area(page_width, page_height, percentage)
              page_text = extract_text_by_area(page, x0, y0, x1, y1).replace('\n', ' ').strip()
              if page_text:
                  extracted_text += page_text + " "
          
          standardized_text = re.sub(r'\s+', ' ', extracted_text).strip()
          return standardized_text
  

  调用时直接传入percentage即可，无需提前获取页面尺寸。

• 增加异常处理：避免文件不存在、页码超出范围、文本提取为空等情况导致脚本崩溃：

  def get_text_with_area_extraction(filepath: str, start_page: int, end_page: int, percentage) -> str:
      if not os.path.exists(filepath):
          raise FileNotFoundError(f"文件 {filepath} 不存在")
      with fitz.open(filepath) as doc:
          if start_page < 0 or end_page >= len(doc) or start_page > end_page:
              raise ValueError("页码范围无效，请检查输入")
          extracted_text = ""
          for page_num in range(start_page, end_page + 1):
              page = doc[page_num]
              page_width = page.rect.width
              page_height = page.rect.height
              x0, y0, x1, y1 = calculate_text_area(page_width, page_height, percentage)
              page_text = extract_text_by_area(page, x0, y0, x1, y1).replace('\n', ' ').strip()
              if page_text:
                  extracted_text += page_text + " "
          
          if not extracted_text:
              raise RuntimeError("未提取到任何文本，请检查clip参数或PDF内容")
          standardized_text = re.sub(r'\s+', ' ', extracted_text).strip()
          return standardized_text
  

• 优化文本清理：增加对页码、特殊符号的清理，提升有声书可读性：

  def clean_text(text):
      # 移除页码（阿拉伯数字或罗马数字）
      text = re.sub(r'\b\d+\b|\b[IVXLCDM]+\b', '', text)
      # 移除多余特殊符号
      text = re.sub(r'[^\w\s.,!?;:\'-]', '', text)
      # 合并连续空格
      text = re.sub(r'\s+', ' ', text).strip()
      return text
  

  在文本提取完成后调用该函数进一步处理。

• 分批生成音频：长文本可能导致gTTS处理失败，可拆分文本分批生成后合并音频（需安装pydub和ffmpeg）：

  from pydub import AudioSegment
  
  def split_text(text, chunk_size=5000):
      # 按句子拆分，避免截断单词
      sentences = re.split(r'(?<=[.!?])\s+', text)
      chunks = []
      current_chunk = ""
      for sentence in sentences:
          if len(current_chunk) + len(sentence) <= chunk_size:
              current_chunk += sentence + " "
          else:
              chunks.append(current_chunk.strip())
              current_chunk = sentence + " "
      if current_chunk:
          chunks.append(current_chunk.strip())
      return chunks
  
  # 使用示例
  text_chunks = split_text(extracted_text)
  audio_segments = []
  for i, chunk in enumerate(text_chunks):
      tts = gTTS(text=chunk, lang='en')
      temp_file = f"temp_chunk_{i}.mp3"
      tts.save(temp_file)
      audio_segments.append(AudioSegment.from_mp3(temp_file))
      os.remove(temp_file)
  
  # 合并音频
  combined_audio = sum(audio_segments)
  combined_audio.export(output_audio_path, format="mp3")
  

内容的提问来源于stack exchange，提问作者Jake DAmico