Python PDF Compress

需求

• 一个PDF预览功能，由于前端预览太大的PDF时加载过慢，想要后端做PDF压缩。
• 对比后基于fitz，将PDF中的每页转换成图片

性能对比

• 源 PDF (Size: 27.9 MB | 3[图片页数] / 8[总页数] )

方法	压缩比例	处理速度	依赖项	缺点
PyPDF2	27.6MB (1%)	0.029s	无	压缩比例低
fitz	7.40MB (27.6%)	2.9s	无
Ghostscript	16MB (27.3%)	5.645s	Ghostscript	需要另裝工具 且需要命令行執行
aspose.pdf	13.5 MB (27.4%)	14.657s	无	有水印
Spire.PDF	6.24 MB (27.8%)	15.17s	无	有水印

实例代码

PyPDF2

import time
from PyPDF2 import PdfReader, PdfWriter


def compress_pdf_pypdf2(input_path, output_path):
    reader = PdfReader(input_path)
    writer = PdfWriter()
    for page in reader.pages:
        writer.add_page(page)
    # 启用内容流压缩（无损）
    writer.compress_content_streams = True
    with open(output_path, "wb") as f:
        writer.write(f)


if __name__ == "__main__":
    start = time.time()
    compress_pdf_pypdf2("input.pdf", "output_pypdf2.pdf")
    print(time.time() - start)

aspose.pdf

• 缺点：商用 - 左上角的水印去不掉

# -*- coding: utf-8 -*-
import time
import aspose.pdf as ap


def Lossless_Compression(path1, path2):
    """
    :param path1: 需要压缩的pdf文件路径
    :param path2: 保存的pdf文件路径
    :return: None
    """
    compress_path = ap.Document(path1)  # 需要压缩的pdf文件路径
    optimize = ap.optimization.OptimizationOptions()
    optimize.image_compression_options.compress_images = True
    optimize.image_compression_options.image_quality = 90  # 压缩质量
    compress_path.optimize_resources(optimize)
    compress_path.save(path2)  # 需要压缩后保存的文件路径


if __name__ == "__main__":
    start = time.time()
    Lossless_Compression("input.pdf", "output_aspose.pdf")
    print(time.time() - start)

aspose.pdf

• 缺点：商用 - 左上角的水印去不掉

import time
from spire.pdf import *


def compress_pdf(input_pdf, output_pdf):
    # 创建 PdfCompressor 对象，并设置压缩选项（设置图片质量、启用图片大小调整和压缩）
    compressor = PdfCompressor(input_pdf)
    compression_options = compressor.OptimizationOptions
    compression_options.SetImageQuality(ImageQuality.Medium)
    compression_options.SetResizeImages(True)
    compression_options.SetIsCompressImage(True)

    # 压缩 PDF 文件并保存到输出文件夹
    compressor.CompressToFile(output_pdf)


if __name__ == "__main__":
    start = time.time()
    compress_pdf("input.pdf", "output_spire.pdf")
    print(time.time() - start)

Ghostscript

• 调用外部工具，压缩比例高，支持多种优化算法。
• 依赖系统安装的Ghostscript，适合对压缩率要求高的场景。

#!/bin/bash

# Compression levels:
#     0: default - almost identical to /screen, 72 dpi images
#     1: prepress - high quality, color preserving, 300 dpi imgs
#     2: printer - high quality, 300 dpi images
#     3: ebook - low quality, 150 dpi images
#     4: screen - screen-view-only quality, 72 dpi images

gs -sDEVICE=pdfwrite \
-dCompatibilityLevel=1.5 \
-dNOPAUSE \
-dQUIET \
-dBATCH \
-sOutputFile=compressed.pdf input.pdf

fitz

import time
import fitz

def auto_dpi(page):
    rect = page.rect
    return max(72, int(300 * (rect.width / 800)))


def compress_pdf(path1, path2, dpi):
    """
    压缩PDF函数（无需中间图片文件）
    :param path1: 原始PDF路径
    :param path2: 输出PDF路径
    :param dpi: 输出分辨率（值越小文件越小，但越模糊）
    """
    # 打开原始PDF
    src_doc = fitz.open(path1)
    # 创建新PDF对象
    dst_doc = fitz.open()

    for page in src_doc:
        # dpi = auto_dpi(page)
        # 生成图片矩阵
        matrix = fitz.Matrix(dpi / 100.0, dpi / 100.0)
        # 获取页面像素图（关闭alpha通道）
        pix = page.get_pixmap(matrix=matrix, alpha=False)

        if page.get_text("text"):
            img_bytes = pix.tobytes(output="png")
        else:
            img_bytes = pix.tobytes(output="jpeg", jpg_quality=85)
        # # 将像素图转为JPEG字节流（内存操作）
        # img_bytes = pix.tobytes(output="jpeg", jpg_quality=90)

        # 从内存字节流创建图片PDF
        imgdoc = fitz.open("jpeg", img_bytes)
        pdfbytes = imgdoc.convert_to_pdf()
        imgdoc.close()

        # 将单页PDF插入目标文档
        pdfpage = fitz.open("pdf", pdfbytes)
        dst_doc.insert_pdf(pdfpage)
        pdfpage.close()

    # 保存并关闭文档
    if dst_doc.page_count > 0:
        dst_doc.save(path2, deflate=True, garbage=3)  # 启用压缩
    dst_doc.close()
    src_doc.close()


if __name__ == "__main__":
    start = time.time()
    zoom = 300  # 缩放比率
    compress_pdf("input.pdf", "output_fitz.pdf", zoom)
    print(time.time() - start)