Standards Development

Scope

The standard will establish requirements for automated systems that capture images of high quality faces that conform to the appearance related requirements of ISO/IEC 19794-5 and ISO/IEC 39794-5 The standard establishes requirements for:

• Face-aware auto-capture: Executing range-measurement, face detection, illumination control, exposure control, and pose estimation (roll, pitch, yaw) in the capture loop.

• Face image quality assessment: Check conformance of an image to quality related clauses of ISO/ICAO portrait requirements.

• High native resolution: Specifications for collection and retention of high resolution photographs preferred by forensic reviewers.

• Conversion of high resolution 2D forensic images into images suited for contemporary face recognition algorithms.

• Capture of 3D information as a supplemental biometric modality and for supporting presentation attack detection.

• Mechanisms to improve exposure (e.g. high dynamic range sensors): To improve imaging across ethnicities and difficult illumination environments.

• Mechanisms for detection and removal of subjects appearing in the background • Compression and resampling: Tightly specify compression for retention of images

• Integrity protection of images: Require cryptographic integrity protection for be applied to the image at source, to impede modification.

• Transmission: Specifications for network connectivity, to elide printing-scanning lifecycle processes.

The standard is primarily intended to support instantiation of fully conformant ISO/IEC 19794-5 and 39794-5 records that will be persist on credentials or in authoritative databases. It may additionally apply to collection of images in authentication or identification attempts also.

The standard might reasonably exclude from its scope requirements on: Form factor; mounting; power consumption; use of particular lenses;

Purpose

This standard is intended to define capture subsystemsX that mitigate the problems enumerated below.

These problems occur with some frequency in

1. Capture of images with zero faces present.

2. Capture of images with multiple people present, some in the background

3. Capture of images with faces present but in which the subject was not ready and looking at, for example, the floor.

4. Capture of images with poor subject behavior (too close to camera, eyes closed etc.)

5. Capture of images with poor imaging properties (inadequate illumination, for example)

6. Capture of images with resolution inadequate for human adjudication.

7. Capture of only frontal views, which are not sufficient for human adjudication.

8. Capture of images for which conformance to ISO/IEC 39794-5 (19794-5) is not documented.

9. Undetected presentation attacks (false negatives), and false positives

10. Tampering of images including morphing, beautification, or other alterations

11. Inappropriate compression, cropping and resizing of images.

12. Imperfect association of images from multiple modalities, collected during the same photo session.

This standard establishes properties for capture subsystems. Its requirements define face-aware capture-subsystems that have knowledge of the appearance of the anatomical object being imaged i.e. the face. Fingerprint sensors and iris cameras function, almost universally, by actively seeking and assessing the image they expect, and are designed to acquire. For face documentary standards exist on how to photograph a face and the properties of the resulting photograph (e.g. portrait, full-frontal view, 128 grey levels), faces are typically acquired without automation, relying on a fixed geometry and environmental control, and photographic competence. It remains common for quality or conformity assessment to either not be done at all or to be de-coupled from the capture process occurring on a client-side computer, or later by a human reviewer (e.g. consular official reviewing a photograph submitted for a visa application).

This proposed standard envisages close to real-time

• assessment of image properties,

• closed-loop adjustment of illumination,

• feedback to subject or photographer.

X The term capture subsystem appears in the harmonized “components of a general biometric system” figure that appears, for example, in CD2 revision 37N6909 of ISO/IEC 19795-1.