A state-of-the-art, enterprise-grade batch OCR processing system that leverages advanced Tesseract capabilities with spatial text reconstruction to preserve document layout, spacing, and formatting. Built with Python's most powerful libraries for high-performance, scalable document processing.
- Layout Preservation: Maintains original document spacing and formatting
- Intelligent Word Positioning: Uses bounding box analysis for accurate text reconstruction
- Table & Column Support: Handles structured documents with proper alignment
- Multi-line Text: Preserves paragraph breaks and text flow
- Multiple Execution Modes: Sequential, Threading, Multiprocessing, and Async
- Auto-scaling Workers: Intelligent resource management based on system capabilities
- Batch Processing: Handle thousands of images efficiently
- Progress Tracking: Real-time processing status with detailed metrics
- PIL Enhancement Pipeline: Contrast, sharpness, and brightness optimization
- OpenCV Integration: Advanced computer vision preprocessing
- Noise Reduction: Gaussian blur and morphological operations
- Adaptive Thresholding: Optimal binary conversion for text extraction
- JSON: Structured data with full metadata and spatial information
- CSV: Tabular format for spreadsheet analysis
- Excel: Multi-sheet reports with summary statistics
- Text Files: Standard (with metadata) and clean (text-only) outputs
- Individual Files: Per-image text extraction with optional metadata
- Combined Files: All results in single files with configurable formatting
- Word-Level Data: Detailed bounding box coordinates for advanced analysis
- Support for 100+ languages via Tesseract
- Configurable language models
- Mixed-language document processing
- Unicode text handling
# Process a single image
python batch_ocr.py document.jpg
# Process multiple images with layout preservation
python batch_ocr.py img1.jpg img2.png img3.tiff
# Process entire directory recursively
python batch_ocr.py /path/to/documents/ -r
# High-performance batch processing
python batch_ocr.py documents/ -m multiprocessing -w 8# Process forms with preserved spacing (default behavior)
python batch_ocr.py forms/ --preserve-layout
# Fine-tune spacing sensitivity for tables
python batch_ocr.py tables/ --space-threshold 40
# Export detailed word-level spatial data
python batch_ocr.py receipts/ --export-word-data# Spanish document processing
python batch_ocr.py documentos/ -l spa
# German documents with advanced preprocessing
python batch_ocr.py dokumente/ -l deu -p opencv
# Mixed language processing
python batch_ocr.py international/ -l eng+spa+fra# Maximum performance with all CPU cores
python batch_ocr.py large_batch/ -m multiprocessing -w $(nproc)
# Memory-optimized for large images
python batch_ocr.py high_res/ -m threading -w 4
# Async processing for I/O intensive workflows
python batch_ocr.py network_storage/ -m async# High-quality preprocessing for poor scans
python batch_ocr.py old_documents/ -p high
# OpenCV preprocessing for difficult images
python batch_ocr.py challenging/ -p opencv
# Verbose logging for debugging
python batch_ocr.py test_batch/ -v
# Quiet mode for automated processing
python batch_ocr.py images/ -q# High-quality preprocessing for poor scans
python batch_ocr.py old_documents/ -p high
# OpenCV preprocessing for difficult images
python batch_ocr.py challenging/ -p opencv
# Verbose logging for debugging
python batch_ocr.py test_batch/ -v┌─────────────────────┐ ┌─────────────────────┐ ┌─────────────────────┐
│ TesseractOCREngine │ │ ImagePreprocessor │ │ BatchOCRProcessor │
│ │ │ │ │ │
│ • Spatial Recon. │ │ • PIL Enhancement │ │ • Multi-threading │
│ • Confidence Calc. │ │ • OpenCV Pipeline │ │ • Multiprocessing │
│ • Layout Analysis │ │ • Noise Reduction │ │ • Async Processing │
└─────────────────────┘ └─────────────────────┘ └─────────────────────┘
│ │ │
└─────────────┬─────────────┘ │
│ │
▼ ▼
┌─────────────────────┐ ┌─────────────────────┐
│ OCRResult │ │ OCRResultExporter │
│ │ │ │
│ • Text Content │ │ • JSON Export │
│ • Spatial Data │ │ • CSV Export │
│ • Confidence │ │ • Excel Export │
│ • Metadata │ │ • Text Files │
└─────────────────────┘ └─────────────────────┘
- Bounding Box Extraction: Extract coordinates for every detected word
- Line Grouping: Group words by vertical position using configurable thresholds
- Horizontal Analysis: Calculate pixel gaps between adjacent words
- Smart Spacing: Insert proportional spaces based on actual distances
- Layout Reconstruction: Rebuild text maintaining original document structure
ocr_output/
├── detailed_ocr_results.json # Full results with spatial data
├── word_level_data.json # Word coordinates and bounding boxes
├── ocr_results.csv # Tabular data for analysis
├── ocr_results.xlsx # Excel with multiple sheets
├── all_extracted_text.txt # Combined text with metadata
├── all_extracted_text_clean.txt # Combined text without metadata
├── clean_text_output.txt # Text-only single file
├── extracted_texts/ # Individual files with metadata
│ ├── document1.txt
│ ├── form2.txt
│ └── receipt3.txt
└── clean_texts/ # Individual files without metadata
├── document1.txt
├── form2.txt
└── receipt3.txt
Source: /path/to/invoice.jpg
Confidence: 94.5%
Processing Time: 2.341s
--------------------------------------------------
INVOICE #12345
Date: 2024-01-15
Amount: $1,250.00
INVOICE #12345
Date: 2024-01-15
Amount: $1,250.00
==================== invoice.jpg ====================
INVOICE #12345
Date: 2024-01-15
Amount: $1,250.00
==================== receipt.jpg ====================
RECEIPT
Store: ABC Market
Total: $45.67
- Enterprise Document Management: Process invoices, contracts, and forms with preserved layout
- Data Extraction: Clean text output for feeding into NLP models and databases
- Archive Digitization: Convert physical documents to searchable digital text
- Compliance & Audit: Maintain document integrity with comprehensive metadata
- Financial Services: Invoice processing, receipt digitization, form automation
- Healthcare: Medical record digitization, prescription processing
- Legal: Contract analysis, case document processing
- Education: Academic paper digitization, assignment processing
- Manufacturing: Quality control documentation, compliance records
- API Workflows: Clean text output for seamless API integration
- Database Import: Direct text content insertion without metadata cleanup
- Content Management: Pure text for CMS systems and search engines
- Analytics Pipelines: Structured data export for business intelligence tools
| Argument | Description | Default | Example |
|---|---|---|---|
input_paths |
Files/directories to process | Required | documents/ |
-r, --recursive |
Process directories recursively | False | -r |
-o, --output-dir |
Output directory | ocr_output |
-o results/ |
-f, --formats |
Output formats | json |
-f json csv xlsx |
-l, --language |
OCR language | eng |
-l spa |
-p, --preprocessing |
Image preprocessing level | medium |
-p opencv |
-m, --mode |
Processing mode | multiprocessing |
-m threading |
-w, --workers |
Number of workers | Auto-detect | -w 8 |
--preserve-layout |
Enable spatial reconstruction | True | --preserve-layout |
--space-threshold |
Spacing sensitivity (pixels) | 30 | --space-threshold 40 |
--export-word-data |
Export word-level coordinates | False | --export-word-data |
--individual |
Create individual text files | False | --individual |
-v, --verbose |
Verbose logging | False | -v |
-q, --quiet |
Suppress output | False | -q |
- Sequential: Single-threaded, minimal memory usage
- Threading: Multi-threaded, I/O bound optimization
- Multiprocessing: Multi-process, CPU intensive tasks
- Async: Asynchronous processing, modern concurrency
- None: No preprocessing, fastest processing
- Low: Basic enhancement
- Medium: Balanced quality/speed (recommended)
- High: Aggressive enhancement for poor quality images
- OpenCV: Advanced computer vision preprocessing
| Language | Code | Installation |
|---|---|---|
| English | eng |
Default |
| Spanish | spa |
sudo apt install tesseract-ocr-spa |
| French | fra |
sudo apt install tesseract-ocr-fra |
| German | deu |
sudo apt install tesseract-ocr-deu |
| Chinese (Simplified) | chi-sim |
sudo apt install tesseract-ocr-chi-sim |
| Japanese | jpn |
sudo apt install tesseract-ocr-jpn |
| Arabic | ara |
sudo apt install tesseract-ocr-ara |
# Process documents with multiple languages
python batch_ocr.py mixed_docs/ -l eng+spa+fra
# Auto-detect language (requires additional setup)
python batch_ocr.py unknown/ -l osd- Word-level confidence: Individual word reliability scores
- Weighted averaging: Character-count based confidence calculation
- Quality filtering: Automatic low-confidence word filtering
- Graceful degradation: Continue processing on individual failures
- Detailed error logs: Comprehensive error reporting
- File integrity: SHA256 hashing for verification
- Format verification: Automatic image format detection
- Dimension capture: Image size and resolution tracking
- Enhancement tracking: Applied preprocessing steps logging
{
"processing_stats": {
"total_files": 150,
"successful_extractions": 148,
"failed_extractions": 2,
"total_processing_time": 45.7,
"average_confidence": 87.3,
"layout_preserved_count": 146
}
}{
"words": [
{
"text": "Invoice",
"confidence": 96,
"bbox": {"left": 100, "top": 50, "width": 120, "height": 25},
"line_num": 1,
"word_num": 1
}
]
}# Ubuntu/Debian
sudo apt install tesseract-ocr
# macOS
brew install tesseract
# Windows: Add to PATH after installation# Try different preprocessing
python batch_ocr.py images/ -p opencv
# Use appropriate language
python batch_ocr.py docs/ -l spa
# Combine both approaches
python batch_ocr.py docs/ -l deu -p high# Reduce workers
python batch_ocr.py large_batch/ -w 2
# Use sequential processing
python batch_ocr.py huge_files/ -m sequential# Adjust spacing threshold for better layout preservation
python batch_ocr.py forms/ --space-threshold 50 --clean-text
# Export word data for manual layout analysis
python batch_ocr.py tables/ --export-word-data --text-only
# Use different output formats for comparison
python batch_ocr.py docs/ --individual --combined --clean-text- Image Quality: Higher DPI generally improves accuracy
- Preprocessing: Match preprocessing level to image quality
- Language Models: Use specific language models when possible
- Worker Count: Optimal workers = CPU cores + 4 for I/O bound tasks
- Batch Size: Process 100-1000 images per batch for optimal performance
- Tesseract OCR for the powerful OCR engine
- OpenCV for advanced image processing capabilities
- Pillow for image manipulation
- PyTesseract for Python integration
Built with ❤️ for the OCR community
Transform your document processing workflows with enterprise-grade OCR capabilities.