The Complete Technical Guide to Passport MRZ Extraction

The Machine Readable Zone (MRZ) represents one of the most standardized document formats in existence. Understanding how to extract and validate MRZ data is essential for any organization processing travel documents.

Understanding MRZ Structure

TD3 Format (Passports)

Standard passports use a two-line MRZ format with 44 characters per line:

Line 1 Structure:

• Position 1: Document type (P = Passport)
• Positions 2-3: Issuing country code (ISO 3166-1 alpha-3)
• Positions 4-44: Surname and given names (separated by <<)

Line 2 Structure:

• Positions 1-9: Passport number
• Position 10: Check digit for passport number
• Positions 11-13: Nationality
• Positions 14-19: Date of birth (YYMMDD)
• Position 20: Check digit for DOB
• Position 21: Sex (M/F/<)
• Positions 22-27: Expiration date
• Position 28: Check digit for expiration
• Positions 29-42: Optional data
• Position 43: Check digit for optional data
• Position 44: Overall check digit

Check Digit Algorithm

MRZ uses a weighted modulo 10 algorithm for validation:

Weights: 7, 3, 1 (repeating)
Character values: 0-9 = 0-9, A-Z = 10-35, < = 0

Example: Validating "AB1234567"
A(10)×7 + B(11)×3 + 1×1 + 2×7 + 3×3 + 4×1 + 5×7 + 6×3 + 7×1
= 70 + 33 + 1 + 14 + 9 + 4 + 35 + 18 + 7
= 191 mod 10 = 1 (check digit)

OCR Challenges and Solutions

Common Recognition Errors

| Character | Often Confused With | Solution | |-----------|-------------------|----------| | 0 (zero) | O (letter) | Context-aware recognition | | 1 (one) | I, L | Font pattern analysis | | 8 | B | Structural analysis | | 5 | S | Curvature detection |

Image Quality Requirements

For optimal extraction accuracy:

• Resolution: Minimum 300 DPI
• Contrast: MRZ area contrast ratio > 70%
• Skew: Less than 15 degrees from horizontal
• Lighting: Even illumination without glare

Advanced Extraction Techniques

Preprocessing Pipeline

1. Grayscale conversion with contrast enhancement
2. Noise reduction using bilateral filtering
3. Binarization with adaptive thresholding
4. Skew correction via Hough transform
5. MRZ zone detection using edge detection and contour analysis

Deep Learning Enhancement

Modern MRZ extraction combines traditional OCR with neural networks:

• CNN-based character recognition improves accuracy on damaged documents
• Sequence-to-sequence models leverage character relationships
• Attention mechanisms focus processing on critical regions

Validation Beyond Check Digits

Complete MRZ validation includes:

1. Format compliance - Character positions match ICAO specifications
2. Check digit verification - All five check digits validate
3. Logical consistency - Birth date precedes expiration, country codes valid
4. Cross-field validation - Name fields match visa/permit if present

Integration Considerations

API Response Structure

{
  "document_type": "PASSPORT",
  "issuing_country": "USA",
  "surname": "SMITH",
  "given_names": "JOHN WILLIAM",
  "passport_number": "123456789",
  "nationality": "USA",
  "date_of_birth": "1985-03-15",
  "sex": "M",
  "expiration_date": "2028-03-14",
  "validation": {
    "all_checks_passed": true,
    "confidence_score": 0.997
  }
}

Error Handling

Implement graceful degradation:

• Return partial results with confidence scores
• Flag specific fields requiring manual review
• Provide image quality feedback for re-capture

Compliance Requirements

Organizations processing MRZ data must comply with:

• GDPR (EU) - Data minimization and purpose limitation
• CCPA (California) - Consumer privacy rights
• PCI-DSS - If processing payment-related identity verification
• ICAO Doc 9303 - Technical standards for machine readable travel documents

Implementing robust MRZ extraction requires attention to both technical accuracy and regulatory compliance. The combination of traditional OCR with modern AI techniques achieves the reliability demanded by border control, financial services, and travel industries.