A powerful TypeScript-based OCR (Optical Character Recognition) tool that extracts text from images, specifically optimized for old newspapers and scanned documents. Built with Bun.js, Tesseract.js, and Sharp.

• 🔍 Multiple OCR methods for best results
• 📰 Optimized for old newspapers and poor-quality scans
• 🖼️ Advanced image preprocessing (contrast enhancement, noise reduction, sharpening)
• 📏 Automatic image upscaling for small images
• 🎯 Multiple page segmentation modes
• 📊 Confidence scoring and method comparison
• ⚡ Fast execution with Bun.js runtime

try-tesseract/
├─ test-images/             # Test images directory (auto-generated during tests)
│  └─ test-image.jpg        # Sample test image
├─ .github/                 # GitHub Actions workflows
│  └─ workflows/
│     └─ ci.yml             # CI/CD pipeline configuration
├─ .gitignore               # Git ignore file
├─ bun.lockb                # Bun lock file for dependencies
├─ eng.traineddata          # Tesseract English language data (auto-downloaded)
├─ ocr.test.ts              # Unit tests for OCR functionality
├─ ocr.ts                   # Main OCR application script
├─ utils.ts                 # Utility functions for image processing
├─ package.json             # Node package configuration and scripts
├─ README.md                # Project documentation (this file)
└─ tsconfig.json            # TypeScript configuration

This application requires Tesseract OCR to be installed on your system. Tesseract.js uses the native Tesseract engine under the hood.

brew install tesseract

sudo apt update
sudo apt install tesseract-ocr
sudo apt install libtesseract-dev

1. Download the installer from GitHub Releases
2. Run the installer and add Tesseract to your PATH
3. Or use Chocolatey:

choco install tesseract

sudo pacman -S tesseract

sudo dnf install tesseract

Verify Installation:

tesseract --version

For more installation details, visit: https://tesseract-ocr.github.io/tessdoc/Installation.html

# macOS/Linux
curl -fsSL https://bun.sh/install | bash

# Windows
powershell -c "irm bun.sh/install.ps1 | iex"

Verify Installation:

bun --version

1. Clone or download this repository

2. Install dependencies:

bun install tesseract.js sharp

3. Save the OCR script as ocr.ts

Extract text with default settings (aggressive preprocessing):

bun run ocr.ts image.png

This will try all preprocessing and segmentation combinations and return the best result:

bun run ocr.ts newspaper.png --all

Use specific page segmentation mode and preprocessing:

# Single column text with aggressive preprocessing (good for newspapers)
bun run ocr.ts newspaper.png --psm 4 --prep aggressive

# Single block with threshold preprocessing
bun run ocr.ts newspaper.png --psm 6 --prep threshold

# Automatic segmentation with basic preprocessing
bun run ocr.ts document.png --psm 3 --prep basic

bun run ocr.ts --help

• PSM 1: Automatic page segmentation with orientation detection
• PSM 3: Fully automatic page segmentation (default)
• PSM 4: Single column of text (best for newspapers)
• PSM 6: Single uniform block of text
• PSM 11: Sparse text detection

• basic: Light enhancement (normalize + sharpen)
• aggressive: Heavy processing for poor quality (contrast boost + noise reduction)
• threshold: Black and white conversion (best for very old documents)

bun run ocr.ts old-newspaper.png --all

Output:

==========================================================
Trying: PSM 3 (Auto) + Aggressive preprocessing
==========================================================
Progress: 100%
Confidence: 87.45%
Characters extracted: 1247

==========================================================
Best method: PSM 4 (Single column) + Basic preprocessing
Confidence: 89.32%
==========================================================

==========================================================
EXTRACTED TEXT:
==========================================================
Union Negro ....
...

bun run ocr.ts document.png --psm 4 --prep basic

bun run ocr.ts poor-quality.png --psm 6 --prep threshold

1. Image Preprocessing: The image is enhanced using Sharp:

   • Upscaling (if needed)
   • Grayscale conversion
   • Contrast normalization
   • Sharpening
   • Noise reduction (optional)
   • Thresholding (optional)

2. OCR Processing: Tesseract analyzes the preprocessed image with specified segmentation mode

3. Result Selection: When using --all, the tool compares all methods and selects the one with the best score (length × confidence)

• Old Newspapers: Use --all or --psm 4 --prep aggressive
• Clean Documents: Use --psm 3 --prep basic
• Poor Quality Scans: Use --prep threshold or --prep aggressive
• Small Images: The tool automatically upscales, but higher resolution inputs work better
• Multiple Columns: Try --psm 3 for automatic detection

• Ensure Tesseract OCR is properly installed
• Verify with tesseract --version
• Check that language data files are installed

bun install sharp

# Make sure Tesseract is installed and working
tesseract --version

# Clean install dependencies
rm -rf node_modules bun.lock
bun install

# Auto-fix most issues
bun run lint:fix
bun run format

• Try --all to find the best method
• Ensure input image is high resolution (at least 300 DPI)
• Try different preprocessing methods
• Check if the image is properly oriented

• Image quality may be too poor
• Try --prep aggressive or --prep threshold
• Consider manually cleaning the image first

The test suite includes:

• ✅ Image preprocessing tests
• ✅ Text extraction tests
• ✅ Multiple PSM mode tests
• ✅ Image format support tests
• ✅ Edge case handling
• ✅ Performance benchmarks

# Run all tests
bun test

# Run with coverage
bun test --coverage

# Run specific test
bun test ocr.test.ts

Note: Tests create temporary images in ./test-images/ directory and clean up automatically.

This project includes a GitHub Actions workflow that automatically:

• ✅ Lints code on every push and pull request
• ✅ Checks formatting to ensure code consistency
• ✅ Runs unit tests to verify functionality
• ✅ Type checks TypeScript code
• ✅ Generates test coverage reports

The CI pipeline runs on:

• Push to main or develop branches
• Pull requests targeting main or develop

Jobs:

1. lint-and-format: Checks code quality and formatting
2. test: Runs unit tests with Tesseract OCR
3. build: Verifies TypeScript compilation
4. coverage: Generates test coverage report (on push only)

To view CI status, check the Actions tab in your GitHub repository.

Add this badge to your README to show CI status:

![CI](https://github.com/rjoydip/try-tesseract/workflows/CI/badge.svg)

• Bun.js - Fast JavaScript runtime
• Tesseract.js - OCR library
• Sharp - Image processing
• Tesseract OCR - OCR engine

Contributions are welcome! Here's how you can help:

1. Fork the repository

2. Clone your fork:

git clone https://github.com/yourusername/try-tesseract.git
cd try-tesseract

3. Install dependencies:

bun install

4. Create a new branch:

git checkout -b feature/your-feature-name

1. Make your changes

2. Run linter and formatter:

bun run lint:fix
bun run format

3. Add tests for new features

4. Run tests:

bun test

5. Commit your changes:

git add .
git commit -m "feat: add your feature description"

6. Push to your fork:

git push origin feature/your-feature-name

7. Create a Pull Request

Follow conventional commits:

• feat: - New feature
• fix: - Bug fix
• docs: - Documentation changes
• test: - Test changes
• refactor: - Code refactoring
• style: - Code style changes (formatting)
• chore: - Build process or auxiliary tool changes

• Use TypeScript strict mode
• Follow the existing code style
• Add JSDoc comments for public functions
• Keep functions small and focused
• Write meaningful variable names

• Write unit tests for all new functions
• Ensure all tests pass before submitting PR
• Aim for high test coverage
• Test edge cases and error conditions

• Keep PRs focused on a single feature/fix
• Update documentation if needed
• Ensure CI checks pass
• Respond to review feedback promptly

MIT

• Tesseract Documentation
• Tesseract.js Documentation
• Sharp Documentation
• Bun Documentation

If you find this project helpful, please consider:

• ⭐ Starring the repository
• 🐛 Reporting bugs
• 💡 Suggesting new features
• 📖 Improving documentation