Engineering a Biometric Photo Compliance Engine for 190+ Countries

The Regulatory Jungle

Most developers assume a passport photo is simply a "square image with a white background."

When researching visa and passport compliance for VisaReadyNow, I spent 4–6 months manually examining immigration regulations across 190+ countries. The reality is wildly fragmented:

• United States: Strict 2x2 inch (51x51mm) square format. The head (from bottom of chin to top of hair) must be between 50% and 69% of the overall image height (1 to 1 3/8 inches). Background must be pure solid white or off-white.
• Schengen Area (EU) / UK / Canada: 35x45mm rectangular format. The head must occupy 70–80% of the vertical frame (32–36mm). The background must be light grey or cream—never pure stark white (which causes edge blending on biometric scanners).
• Malaysia: 35x50mm with a solid light blue background.
• DPI Metadata: File uploads are rejected by consular validation systems if the EXIF DPI header is missing or uncalibrated to 300/600 DPI.

Extracting the Rules into Open-Source Code

Rather than hardcoding conditional statements throughout our application codebase, I normalized all national requirements into a typed specification schema.

On August 31, 2026, I released this specification database to the community across both ecosystems:

• npm (TypeScript): visareadynow-passport-specs
• PyPI (Python): visareadynow-passport-specs

export interface CountryPassportSpec {
  countryCode: string;
  countryName: string;
  widthMm: number;
  heightMm: number;
  minHeadHeightPercent: number;
  maxHeadHeightPercent: number;
  backgroundColor: 'white' | 'light-grey' | 'light-blue' | 'off-white';
  recommendedDpi: number;
  printableGrid: {
    sheetSize: '4x6' | 'A4';
    photosPerSheet: number;
  };
}

The Facial Landmark Pipeline

1. Detection & Bounding Box: Detect face bounding coordinates, chin base, and top-of-skull landmarks.
2. Dynamic Scaling: Calculate the required crop bounds such that (chin - topOfHead) / totalHeight falls strictly within the target country's statutory percentage range.
3. Background Replacement: Apply semantic segmentation to separate hair, ears, and shoulders, blending edges onto the country's exact RGB specification.
4. Printable Sheet Assembly: Automatically arrange 4 or 6 copies on a 4x6 inch canvas with precise cut lines for pharmacy printing booths.