Facial recognition standards

October 2008 Access Control & Identity Management

Facial recognition only works when images all adhere to ISO/IEC and ICAO international standards.

Facial recognition is a chosen biometric technology for solutions such as e-Passport and e-Visa applications. Currently it is also being introduced in national ID schemes as well as border control watch-list surveillance systems all over the world. Moreover, access control applications using biometric facial verification are being introduced on a daily basis in many different environments.

The success of a facial recognition system will depend on several factors that influence the quality of the image. The acquisition process (including the capturing device such as a camera) as well as the biometric characteristics of a person play a major role in the performance of a facial recognition system.

ISO/IEC biometric standards provide guidelines for face image capturing as well as image quality. Additionally ICAO (International Civil Aviation Organization) specifies guidelines on how photographs should be taken for travel documents.

Recommendations by ISO/IEC and ICAO on how to take photographs for e-Passports, e-Visas or ID documents are as follows:

Full frontal facial image: this type of image includes the full head with all hair in most cases, as well as neck and shoulders. This type is suitable for storage of the facial info, and is applicable to portraits for passport, driver licence and ‘mugshot’ images.

Token facial image: an image type that specifies frontal images with a specific geometric size and eye positioning based on the width and height of the image. This image is suitable for minimising the storage requirements for computer face recognition tasks such as verification, while still offering vendor independence and human verification capabilities.

Factors that affect the facial recognition systems performance are categorised as concerning the environment and the user (subject). In the acquisition and capture process of a face, elements such as variation of lighting and extreme or weak illuminations as well as camera characteristics (sensor resolution) play a role in performance. Current facial recognition standards for e-Passports allow for limited facial expressions and behaviour like closed eyes and subject position.

In the analysis of facial image quality different aspects have to be considered

Image resolution and size: for the full frontal image type recommended standard (ISO/IEC) is 180 pixels of resolution of the width of the head or roughly 90 pixels from the eye to the centre.

Image exposure or noise

Different processes that are required in producing a digital image contribute to noise in facial images. Noise can be generated by the CCD sensor of a camera, image scanning device or image compression algorithm (eg, JPEG).

Lighting and background: The best results to enhance machine-assisted face recognition performance are achieved when photographs are taken with a background with a grey plain smooth surface.

No unnatural colour should be introduced. Grayscale photographs should be produced from common incandescent light sources. Colour balancing techniques should be used for colour photographs (high colour-temperature flask with standard film or tungsten-balanced film with incandescent lighting).

The behaviour of the person

Examples of characteristics related to the person’s behaviour include the following:

* Behaviour regarding eyes, eg, closed eyes.

* Closed or open mouth.

* Any kind of expression, eg, smiling or neutral.

* Head pose, eg, frontal or rotated in any direction.

Best practices for face image data

Photographic quality requirements:

* Close-up of your head and shoulders so that your face takes up 70-80% of the photograph.

* In sharp focus and clear.

* Show you looking directly into the camera.

* Show your natural skin tones.

* Have appropriate brightness and contrast.

Photograph style and lighting must:

* Be colour neutral.

* Show your eyes open and clearly visible – no hair across your eyes.

* Be taken with a plain light-coloured background.

* Be taken with uniform lighting and not show shadow or flash reflections on your face.

* No red eyes.

If subject wears glasses:

* The photograph must show eyes clearly with no flash reflection off the glasses.

* No tinted lenses, lighter framed glasses are preferred.

If wearing head coverings:

* These are not permitted except for religious reasons, but your facial features from bottom of chin to top of forehead and both edges of your face must be clearly shown.

The above specifications can be used as guidelines in other applications such as taking photographs of new and existing employees for the purpose of access control ID cards, and later for use in facial recognition search engines for identification of persons.

For more information contact Basia Nasiorowska, Interoptic Networks, [email protected]





Share this article:
Share via emailShare via LinkedInPrint this page



Further reading:

New State of Physical Access Control Report from HID
HID Global Editor's Choice Access Control & Identity Management News & Events
HID released the 2024 State of Physical Access Control Report, identifying five key trends shaping access control's future and painting a picture of an industry that has been undergoing considerable transformation.

Read more...
Smart intercoms are transforming access control
Access Control & Identity Management Products & Solutions
Smart intercoms have emerged as a pivotal tool in modern access control. They provide a seamless and secure way to manage entry points without the need for traditional security guards to validate visitors before granting them access.

Read more...
Easy, secure access for student apartments
Paxton Access Control & Identity Management Surveillance
Enhancing Security and Convenience at Beau Vie II Student Accommodation, a student apartment block located at Banghoek Road, Stellenbosch, with Paxton's access control and video management solution

Read more...
Invixium acquires Triax Technologies
News & Events Access Control & Identity Management
Invixium has announced it has acquired Triax Technologies to expand its biometric solutions with AI-based RTLS (Real-Time Location Systems) offering for improved safety and productivity at industrial sites and critical infrastructure.

Read more...
ControliD's iDFace receives ICASA certification
Impro Technologies News & Events Access Control & Identity Management
The introduction of Control iD's iDFace facial biometric reader, backed by mandatory ICASA certification, underscores the commitment to quality, compliance, and innovation.

Read more...
The future of workplace access
HID Global Access Control & Identity Management
Mobile credentials are considerably more secure than physical access control, because they eliminate the need for physical cards or badges, support multiple security protocols, and add layers of protection on top of basic card encryption.

Read more...
Integrated, mobile access control
SA Technologies Entry Pro Technews Publishing Access Control & Identity Management
SMART Security Solutions spoke to SA Technologies to learn more about what is happening in the estate access world and what the company offers the residential estate market.

Read more...
Bespoke access for prime office space
Paxton Access Control & Identity Management Residential Estate (Industry)
Nicol Corner is home to a six-star fitness club, prime office space, and an award-winning rooftop restaurant. It is also the first building in South Africa to have its glass façade fully incorporate fritted glazing, saving 35% on energy consumption.

Read more...
Next-generation facial recognition access control system
Enkulu Technologies Products & Solutions Access Control & Identity Management Residential Estate (Industry)
With a modern and innovative design, iDFace is the ideal device for monitoring and controlling people entering and exiting a building using facial recognition technology, including liveness detection, for enhanced security.

Read more...
Long-distance vehicle identification
STid Security Products & Solutions Access Control & Identity Management Residential Estate (Industry)
The STid SPECTRE reader can identify vehicles up to 14 metres away, across four traffic lanes, ensuring secure access to an estate without disrupting the traffic flow.

Read more...