利用python将图片版PDF转文字版PDF

pip install PyPDF2
pip install ghostscript
pip install baidu-aip
pip install pdfkit

cd 下载目录
pip install PythonMagick‑0.9.13‑cp36‑cp36m‑win_amd64.whl

#导入PdfFileReader和PdfFileWriter
from PyPDF2 import PdfFileReader, PdfFileWriter
#获取一个pdf对象
pdf_input = PdfFileReader(open(r'pdf路径', 'rb'))
#获取pdf页数
page_count = pdf_input.getNumPages()
#获取pdf第四页的内容
page = pdf_input.getPage(3)
page['/Contents']
#获取一个pdfWriter对象
pdf_output = PdfFileWriter()
# 将一个 PageObject 加入到 PdfFileWriter 中
pdf_output.addPage(page)
#把新pdf保存
pdf_output.write(open(r'新pdf路径','wb'))

#导入baidu-aip
from aip import AipOcr
#https://console.bce.baidu.com/#/index/overview
#产品服务->人工智能->文字识别->创建应用
#获取以下三个值
APP_ID = '??'
API_KEY = '??'
SECRET_KEY = '?? '
#新建一个AipOcr
client = AipOcr(APP_ID, API_KEY, SECRET_KEY)
#读取本地图片的函数
def get_file_content(filePath):
  with open(filePath, 'rb') as fp:
    return fp.read()
#读取本地图片
image = get_file_content('p1.jpg')
#可选参数
options = {}
options["language_type"] = "CHN_ENG"
options["detect_direction"] = "true"
options["detect_language"] = "true"
options["probability"] = "true"
#通用文字识别
client.basicGeneral(image, options)

#读取网络图片
url = "https://note.youdao.com/yws/public/resource/1577071c1ffa2b6bf4e238ef6dbcfbf5/xmlnote/E5A19BEDFEBA4879B217C5BBF53B0245/22138"
#可选参数
options = {}
options["language_type"] = "CHN_ENG"
options["detect_direction"] = "true"
options["detect_language"] = "true"
options["probability"] = "true"
#通用文字识别
client.basicGeneralUrl(url, options)

#读取本地表格图片的函数
def get_file_content(filePath):
  with open(filePath, 'rb') as fp:
    return fp.read()
#读取本地表格图片
image = get_file_content('p2.jpg')
#可选参数
options = {}
options["language_type"] = "CHN_ENG"
options["detect_direction"] = "true"
options["detect_language"] = "true"
options["probability"] = "true"
#通用文字识别
client.basicGeneral(image, options)
#读取表格分割效果较差!

#pdfkit安装位置设置
path_wk = r'pdfkit安装位置设置'
pdfkit_config = pdfkit.configuration(wkhtmltopdf = path_wk)
#pdfkit参数
pdfkit_options = {'encoding': 'UTF-8',}
#制作PDF
pdfkit.from_string(('string'),'D:\test.pdf',configuration=pdfkit_config,options=pdfkit_options)

#导入所需包

#os,操作文件和路径
import os
#ghostscript,代码简化
import ghostscript
#pypdf2,拆分pdf
from PyPDF2 import PdfFileReader, PdfFileWriter
#PythonMagick,单页PDF转图片
from PythonMagick import Image
#baidu-aip,百度文字识别
from aip import AipOcr
#pdfkit,字符串制作PDF
import pdfkit

#参数

path='??'
pdfname='??'
DPI='85'
#https://console.bce.baidu.com/#/index/overview
#产品服务->人工智能->文字识别->创建应用
#获取以下三个值
APP_ID = '??'
API_KEY = '??'
SECRET_KEY = '?? '
#pdfkit安装位置设置
path_wk = r'pdfkit安装位置设置'
pdfkit_config = pdfkit.configuration(wkhtmltopdf = path_wk)
#pdfkit参数
pdfkit_options = {'encoding': 'UTF-8',}

#PDF转化为图片

os.chdir(path)
pdf_input = PdfFileReader(open(pdfname, 'rb'))
#自动获取PDF页数
page_count = pdf_input.getNumPages()
page_range=range(page_count)
#也可以手工指定PDF需要转换的页数
#page_range=range(0,100)
#使用PyPDF和ghostscript
#==超级好用,超级直观,超级短==
for page_num in page_range:
  im = Image()
  im.density(DPI)
  im.read(pdfname + '[' + str(page_num) +']')
  im.write(str(page_num)+ '.jpg')

#图片转化为字符串

#新建一个AipOcr
client = AipOcr(APP_ID, API_KEY, SECRET_KEY)
#读取本地图片的函数
def get_file_content(filePath):
  with open(filePath, 'rb') as fp:
    return fp.read()
#可选参数
options = {}
options["language_type"] = "CHN_ENG"
options["detect_direction"] = "false"
options["detect_language"] = "false"
options["probability"] = "false"
allteststr=[]
for page_num in page_range:
  #读取本地图片
  image = get_file_content(r'%s\%s.jpg' % (path,page_num))
  #通用文字识别,得到的是一个dict
  testjson=client.basicGeneral(image, options)
  teststr=''
  for x in testjson['words_result']:
    teststr=teststr+x['words']+'</br>'
  allteststr.append(teststr)

#字符串写入PDF
for page_num in page_range:
  pdfkit.from_string((allteststr[page_num]),'%s.pdf' % (str(page_num)),configuration=pdfkit_config,options=pdfkit_options)
#合并单页PDF
pdf_output = PdfFileWriter()
for page_num in page_range:
  os.chdir(path)
  pdf_input = PdfFileReader(open('%s.pdf' % (str(page_num)), 'rb'))
  page = pdf_input.getPage(0)
  pdf_output.addPage(page)
pdf_output.write(open('newpdf.pdf','wb'))