PDF 压缩

JavaScript pdf-lib 重编码内嵌 JPEG（浏览器内，不上传） 复制

import { PDFDocument, PDFName, PDFNumber, PDFRawStream } from 'pdf-lib';

const doc = await PDFDocument.load(bytes, { ignoreEncryption: true });
const ctx = doc.context;
for (const [ref, obj] of ctx.enumerateIndirectObjects()) {
  if (!obj?.dict) continue;
  if (String(obj.dict.lookup(PDFName.of('Subtype'))) !== '/Image') continue;
  if (String(obj.dict.lookup(PDFName.of('Filter'))) !== '/DCTDecode') continue;
  if (obj.dict.has(PDFName.of('SMask'))) continue;      // 带透明通道的跳过
  const jpeg = await reEncode(obj.contents, 0.55, 1600); // canvas 解码→限边长→重压
  obj.dict.set(PDFName.of('Width'), PDFNumber.of(jpeg.w));
  obj.dict.set(PDFName.of('Height'), PDFNumber.of(jpeg.h));
  ctx.assign(ref, PDFRawStream.of(obj.dict, jpeg.bytes));
}
const smaller = await doc.save();

Python pikepdf + Pillow 压缩图像流 复制

import io
from PIL import Image
import pikepdf

pdf = pikepdf.open('input.pdf')
for page in pdf.pages:
    for name, image in list(page.Resources.get('/XObject', {}).items()):
        if image.get('/Filter') == '/DCTDecode':
            im = Image.open(io.BytesIO(image.read_bytes()))
            im.thumbnail((1600, 1600))
            buf = io.BytesIO()
            im.convert('RGB').save(buf, 'JPEG', quality=55)
            pdf.objects[image].write(buf.getvalue())     # 原样替换 JPEG 流
pdf.save('output.pdf')