Return to the SDK contents and list of components
Face Image Processing component
The Face Image Processing component creates face templates from face images and is designed to be used in high-volume server applications, which run on server hardware with at least Intel Xeon Gold 6126 (2.6 GHz) processor. The component performs template extraction at a speed of 3,000 faces per minute.
The component also allows to integrate JPEG 2000 with Lossy and Lossless Face Profiles support into systems based on MegaMatcher SDK.
Device Manager software allows to perform simultaneous capture from multiple cameras. Integrators can write plug-ins to support their cameras or other devices using the plug-in framework provided with the Device Manager.
The Face Image Processing component also includes proprietary algorithms, which provide these advanced functionalities after facial template extraction:
- Person's gender recognition.
- Emotions detection: confidence values returned for neutral mood, anger, disgust, fear, happiness, sadness and surprise.
- Facial feature points extraction for each person from an image.
- Age estimation for each person from an image.
- Additional face attributes detection: smile, open-mouth, blink (closed-eyes), glasses, dark-glasses, beard and mustache.
Face liveness detection can be used for determining whether a face in a video stream belongs to a real human or is a photo.
See recommendations for face liveness detection for more information.
- Up to 5 video streams can be processed in parallel with this component. Web Service component from the Face Verification system can be added to the system for perfoming face liveness check with a larger number of video streams.
Captured faces can be checked for compliancy with ICAO requirements. These requirements are checked:
- image pixelation, washed out colors;
- background uniformity (any background can be replaced with constant or transparent automatically);
- face darkness, skin tone, skin reflections, glasses reflections;
- red eyes, looking away eyes (the red eyes can be corrected automatically).
The Face Image Processing component can provide token face images * compatible with the Face Image Format as in ISO/IEC 19794 standard. This face image format enables range of applications on variety of devices, including devices that have limited resources required for data storage, and improves recognition accuracy by specifying data format, scene constraints (lighting, pose), photographic properties (positioning, camera focus) and digital image attributes (image resolution, image size). The following features are available:
- Face Token Image creation from an image containing human face using eye coordinates which may be either hand marked or detected automatically using Neurotechnology face detection algorithm.
- Face is detected and eye coordinates are acquired using state-of-the-art Neurotechnology face detection and recognition algorithm.
- Geometrical face image normalization according to the proportions and photographic properties, which are specified in ISO/IEC 19794 standard.
- Intelligent image padding algorithm for cutting off parts of token face image as specified in ISO/IEC 19794 standard.
Evaluation of the created token face image for the following quality criteria suggested in ISO/IEC 19794 standard:
- Background uniformity – the background in the token face image should be uniform, not cluttered.
- Sharpness – the token face image should not be blurred.
- Too light or too dark images – the token face image should not be too dark or too light.
- Exposure range of an image – the token face image should have a reasonable exposure range to represent as much details of the subject in the image as possible.
- Evaluation of the token face image quality based on suggestions of ISO/IEC 19794 standard (using the quality criteria above).
The Face Image Processing component allows to integrate support for facial image format standards with new or existing biometric systems based on MegaMatcher SDK. These biometric standards are supported:
- BioAPI 2.0 (ISO/IEC 19784-1:2006) (Framework and Biometric Service Provider for Face Identification Engine)
- CBEFF V1.2 (ANSI INCITS 398-2008) (Common Biometric Exchange Formats Framework)
- CBEFF V2.0 (ISO/IEC 19785-1:2006 with Amd. 1:2010, 19785-3:2007 with Amd. 1:2010) (Common Biometric Exchange Formats Framework)
- CBEFF V3.0 (ISO/IEC 19785-3:2015) (Common Biometric Exchange Formats Framework)
- ISO/IEC 19794-5:2005 (Biometric Data Interchange Formats - Face Image Data)
- ISO/IEC 19794-5:2011 (Biometric Data Interchange Formats - Face Image Data)
- ANSI/INCITS 385-2004 (Face Recognition Format for Data Interchange)
- ANSI/NIST-CSL 1-1993 (Data Format for the Interchange of Fingerprint, Facial, & SMT Information)
- ANSI/NIST-ITL 1a-1997 (Data Format for the Interchange of Fingerprint, Facial, & SMT Information)
- ANSI/NIST-ITL 1-2000 (Data Format for the Interchange of Fingerprint, Facial, & SMT Information)
- ANSI/NIST-ITL 1-2007 (Data Format for the Interchange of Fingerprint, Facial, & Other Biometric Information)
- ANSI/NIST-ITL 1a-2009 (Data Format for the Interchange of Fingerprint, Facial, & Other Biometric Information)
- ANSI/NIST-ITL 1-2011 (Data Format for the Interchange of Fingerprint, Facial, & Other Biometric Information)
- ANSI/NIST-ITL 1-2011 Update:2013 Edition 2 (Data Format for the Interchange of Fingerprint, Facial, & Other Biometric Information)
- ANSI/NIST-ITL 1-2011 Update:2015 (Data Format for the Interchange of Fingerprint, Facial, & Other Biometric Information)
One Face Image Processing component license is included with MegaMatcher 13.0 Standard SDK and MegaMatcher 13.0 Extended SDK. The license can be used on Microsoft Windows or Linux x86-64 platform. More licenses for this component can be purchased any time by MegaMatcher 13.0 SDK customers.
Note: Token in this context is used as "symbolic image, good enough image for machine recognition." Token Image as in ISO/IEC19794-5: "A Face 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 type is suitable for minimizing the storage requirements for computer face recognition tasks such as verification while still offering vendor independence and human verification (versus human examination which requires more detail) capabilities."