OCRex

OCRex is a high-accuracy, non-destructive bulk PDF OCR processing tool built on OCRmyPDF. Designed for large-scale document analysis, it enhances low-quality declassified PDFs with advanced pre-processing such as automatic rotation, deskewing, and denoising. The tool leverages multiprocessing and robust error handling to ensure high performance even on low-end machines.

---

Features

• Batch OCR Processing – Processes hundreds of PDFs concurrently.
• Non-Destructive – Preserves original PDFs by generating new output files.
• Advanced Pre-Processing – Automatically rotates, deskews, and denoises scanned pages.
• Multiprocessing – Optimized for high-speed processing using available CPU cores.
• Full OCRmyPDF Integration – Utilizes all features of OCRmyPDF for lossless text recognition.
• Cross-Platform – Works on Linux, macOS, and Windows.

---

Installation

1. Install Dependencies

pip install -r requirements.txt

For system dependencies:

sudo apt install tesseract-ocr ghostscript poppler-utils -y  # Debian/Ubuntu
brew install tesseract ghostscript poppler                   # macOS

2. Install from Source

git clone https://github.com/sbl8/ocrex.git
cd ocrex
python setup.py install

3. Running from Source

python -m ocrex.main --input_dir ./pdfs --output_dir ./output

---

Building a Standalone Binary

pip install pyinstaller
pyinstaller --onefile --name ocrex ocrex/main.py
mv dist/ocrex /usr/local/bin/

Now you can run:

ocrex --input_dir ./pdfs --output_dir ./output

---

Usage Examples

Standard OCR Processing

ocrex --input_dir ./pdfs --output_dir ./output

Correcting Skewed or Rotated Pages

ocrex --input_dir ./pdfs --output_dir ./output --enable_preprocessing --auto_rotate --deskew --denoise

Optimize Speed (More CPU Cores)

ocrex --input_dir ./pdfs --output_dir ./output --workers 8

Optimize File Size

ocrex --input_dir ./pdfs --output_dir ./output --optimize_level 2

---

Workflow for Declassified Documents

Step 1: Pre-Processing for Low-Quality Scans

Use advanced pre-processing to correct skewed, rotated, or noisy scans:

ocrex --input_dir ./pdfs --output_dir ./output --enable_preprocessing --deskew --denoise --auto_rotate

Step 2: OCR Processing with Enhanced Text Retention

Run OCR with a lower optimization level to preserve as much text as possible:

ocrex --input_dir ./pdfs --output_dir ./output --optimize_level 0

Step 3: Verifying OCR Output

Search for key phrases in the OCR output to verify accuracy:

grep "classified" ./output/*.pdf

---

Command-Line Options

Run the following to view all options:

ocrex --help

Option	Description
--input_dir	Path to input PDFs (required).
--output_dir	Path to save OCR PDFs.
--workers	Number of parallel processes (default: CPU count).
--auto_rotate	Automatically rotate misaligned pages.
--deskew	Enable deskewing (default: on).
--denoise	Enable denoising (default: on).
--optimize_level	OCRmyPDF optimization level (0-3, default: 1).
--pdf_dpi	DPI for PDF-to-image conversion (default: 300).
--verbose	Enable verbose logging.

---

Automated Testing & CI/CD

• Tests Directory: All tests are located in the tests/ directory at the project root.
• Continuous Integration: GitHub Actions workflows run tests on every push and pull request.
• Automated Releases: Standalone binaries and source releases are built and tagged automatically via GitHub Actions.

---

License

Distributed under the MIT License. See LICENSE.

---

Contributing

Contributions, bug reports, and feature requests are welcome. Please open an issue or submit a pull request on GitHub.