NIST updates biometric image quality assessment evaluation, seeks feedback

Image quality assessment algorithms are getting better at detection specific defects that will impact face biometrics match success, according to a new report from the U.S. National Institute of Standards and Technology.

NIST’s Face Analysis Technology Evaluation (FATE) Part 11 report evaluates algorithms for Face Image Quality Vector Assessment, in terms of their performance for Quality Specific Image Defect Detection (SIDD).

The latest evaluation adds a pair of submissions from secunet, as well as modified measures for subjects with open eyes and mouths, and occluded faces. The measurement modifications were made to better align the test with the specifications of ISO/IEC 29794-5:2024.

Prior submissions have been entered by Digidata, FRP LLC, Neurotechnology, Rank One Computing, Fraunhofer IGD, Idemia, Dermalog and Seamfix, along with two prior submissions from secunet.

All of the 13 “algorithms submitted have some success at measuring various quality-related parameters.”

The evaluation evaluated a use case involving machine-readable travel documents, how long it takes to produce results, and how accurately algorithms count the number of faces in a frame. Assessments of yaw, pitch and roll angle, as well as background uniformity, resolution, and under- and overexposure are also evaluated. Algorithms were also evaluated for detection of eyeglasses and sunglasses, and motion blurring.

Algorithms from secunet delivered a particularly sharp decrease in false non-match rate (FNMR) as its quality assessments were used to discard lower-quality images.

Comments are invited through December 14.

Article Topics

biometric data quality  |  biometric matching  |  biometric testing  |  biometrics  |  Face Analysis Technology Evaluation (FATE)  |  facial recognition  |  NIST  |  secunet