Overview
ISO/IEC TR 29794-5:2010 - "Information technology - Biometric sample quality - Part 5: Face image data" is an ISO/IEC Technical Report that defines concepts, terminology and recommended approaches for assessing face image quality in the context of automated face recognition systems. It explains the purpose and interpretation of face quality scores (FQS) and describes how face quality assessment algorithms (FQAA) can produce objective, quantitative quality metrics. The report does not standardize quality algorithms nor prescribe performance test methods.
Key topics
- Terms and definitions relevant to face image quality, including comparison score, Face Quality Assessment Algorithm (FQAA), Face Quality Score (FQS), Quality Score Normalization (QSN) and related abbreviations.
- Approaches to face image quality: defining quality relative to automated face recognition performance and describing how quality measures can guide preprocessing, matcher selection or thresholding.
- Categorization framework separating factors that affect quality:
- Subject characteristics (static vs dynamic) - anatomical features, pose, expression, occlusion (e.g., hair or glasses), eye closure.
- Acquisition process properties (static vs dynamic) - sensor/camera characteristics, resolution, compression, illumination, background motion.
- Facial image quality analysis covering measurable image properties and appearance attributes:
- Image size and resolution, noise, contrast, brightness, exposure, focus/blur, color, and subject-camera distance.
- Dynamic subject behaviours and asymmetries (e.g., head pose, facial expression) and scene-level factors (uneven lighting, moving objects).
- Quality score generation concept: atomic FQSs for individual characteristics are fused into a final quality score intended to predict recognition metrics (false match/false non-match risk) without prescribing a single algorithm.
Practical applications and users
ISO/IEC TR 29794-5 is useful for:
- Biometric system designers and integrators seeking to incorporate automated face quality assessment into pipelines (enrollment, live capture, or video).
- Algorithm developers and researchers designing FQAAs and interpreting FQS outputs.
- Test labs and evaluators that need a common vocabulary and classification for face image defects and quality attributes.
- Government agencies, ID/credential issuers, and vendors implementing face recognition for access control, border control, e-passports or surveillance who require guidance on what image attributes affect recognition performance.
Related standards
- ISO/IEC 29794-1 (Framework for biometric sample quality)
- ISO/IEC 19794-5 (Face image data - data interchange format) - provides complementary scene and photographic specifications referenced by this report.
Keywords: ISO/IEC TR 29794-5:2010, face image quality, biometric sample quality, face image data, face quality score, FQAA, biometric systems, face recognition, image resolution, illumination, exposure, focus, occlusion.