Every time you apply for an identity document — a passport, a national identity card, a residence permit — you submit a photograph. Most people give this little thought beyond ensuring the photo is recent and reasonably flattering. But behind that photograph lies a detailed and legally binding technical specification, and whether your image meets it has direct consequences for the reliability of every biometric check performed against your document for the next ten years.
The standard that governs your passport photo
ISO/IEC 39794-5 defines the technical requirements for face images stored in biometric identity documents. Its requirements cover not just the obvious — frontal pose, neutral expression, eyes open — but dozens of technical properties: background uniformity, illumination consistency, the precise inter-eye distance as a proportion of image width, margins around the face, and absence of shadows across facial features.
These requirements exist because face recognition algorithms depend on them. An image that fails to meet pose requirements — even by a few degrees of yaw or pitch — provides a suboptimal template for recognition, increasing both false rejection rates and potentially the vulnerability to impostor attempts.
The compliance gap in remote issuance
The challenge is most acute in remote identity issuance scenarios — where an applicant submits their own photograph through an online channel. Research within the EINSTEIN project — reported in the BioGaze paper — has directly evaluated automated quality assessment tools against the photographic requirements of ISO/IEC 39794-5, examining how well they can detect non-compliant images and provide actionable feedback to applicants.
What actionable feedback looks like
A quality assessment system that tells an applicant their image does not meet requirements is of limited value. One that tells them their head is slightly tilted to the left and they should straighten their pose and retake is actionable. Component-level quality measures, as defined in ISO/IEC 29794-5 and implemented in tools like OFIQ, provide exactly this granularity — each of the 28 quality components can fail independently, and the specific pattern of failures tells both the applicant and the reviewing system exactly what needs to be corrected.
© 2026 EINSTEIN Consortium. EINSTEIN is funded by the European Union’s Horizon Europe programme (GA No. 101121280) and by UKRI (IFS 10093453). Views expressed are those of the authors only. www.einstein-horizon.eu