full-stack-skills OCRmyPDF — Image Processing Guide
Overview
OCRmyPDF includes powerful image processing capabilities to improve scan quality before OCR. These tools help fix skewed pages, remove noise, clean borders, and enhance readability.
For core OCR functionality, see the ocrmypdf skill. For optimization and PDF/A options, see ocrmypdf-optimize. For batch/Docker/scripting, see ocrmypdf-batch.
Deskew
Deskew corrects pages that are slightly rotated (e.g., from feed scanner skew).
# Auto deskew (recommended)
ocrmypdf --deskew input.pdf output.pdf
# Force deskew even if rotation is minimal
ocrmypdf --deskew --force-ocr input.pdf output.pdf
Rotation
Rotate pages to correct upside-down or sideways scans:
# Auto-rotate based on text orientation
ocrmypdf --rotate-pages input.pdf output.pdf
# Force rotate all pages
ocrmypdf --rotate-pages --force-ocr input.pdf output.pdf
Remove Borders / Cleaning
Remove unwanted borders, artifacts, and noise from scanned pages:
# Remove borders (dots, solid borders)
ocrmypdf --remove-bordering input.pdf output.pdf
# Combine with cleanup
ocrmypdf --remove-bordering --clean input.pdf output.pdf
Despeckle
Remove speckles and isolated noise pixels:
# Remove speckles
ocrmypdf --despeckle input.pdf output.pdf
# Aggressive despeckle for very noisy scans
ocrmypdf --despeckle --clean input.pdf output.pdf
Unpaper
unpaper provides advanced post-processing:
# Apply unpaper with default settings
ocrmypdf --unpaper input.pdf output.pdf
# Custom unpaper board options
ocrmypdf --unpaper-args "--board A4" input.pdf output.pdf
Oversampling
Increase image resolution before OCR for better accuracy:
# Oversample to 300 DPI before OCR
ocrmypdf --oversample 300 input.pdf output.pdf
# Common for low-resolution scans
ocrmypdf --oversample 400 input.pdf output.pdf
Combined Recipes
Fix a skewed scan
ocrmypdf --deskew --remove-bordering --despeckle scanned.pdf fixed.pdf
Clean up a very noisy scan
ocrmypdf --deskew --rotate-pages --despeckle --clean --oversample 300 noisy.pdf clean.pdf
Remove all artifacts
ocrmypdf --remove-bordering --unpaper --despeckle dirty.pdf clean.pdf
Quick Reference
| Task | Command |
|---|---|
| Auto deskew | --deskew |
| Auto rotate | --rotate-pages |
| Remove borders | --remove-bordering |
| Remove speckles | --despeckle |
| Unpaper | --unpaper |
| Oversample DPI | --oversample N |
Troubleshooting
- Poor OCR after cleaning: Try
--oversample 300to increase input quality. - Artifacts remain: Use
--unpaperfor aggressive cleanup. - Over-cleaned image: Reduce cleaning options for preserve original quality.
国内适配
- 支持中文文档和中文注释
- 示例代码兼容国内开发环境
- 提供中文 FAQ 和常见问题解答
能力边界
✅ 适用场景
- 当你需要使用此技能对应的技术栈时
- 当项目需要遵循最佳实践时
- 当需要快速上手或深入理解核心概念时
⚠️ 需要注意
- 复杂业务逻辑需要结合具体场景调整
- 性能优化需要根据实际数据量评估
❌ 不适用场景
- 不相关的技术栈或框架
- 需要完全自定义的特殊场景
使用流程
Step 1: 环境准备
确保开发环境已安装必要的依赖和工具。
Step 2: 配置初始化
根据项目需求进行基础配置。
Step 3: 核心功能使用
按照示例代码实现核心功能。
Step 4: 测试验证
运行测试确保功能正常。
Step 5: 部署上线
完成开发后进行部署和监控。