Daon, Pixel in image quality pole positions as paused NIST biometric tests resume

Biometric image quality analysis algorithms are getting better at predicting when facial recognition will fail, according to evaluations by America’s technology testing authority, but gradually. Some of the best on the market were submitted for evaluation by the National Institute of Standards and Technology several years ago

Assessments of algorithms for face and iris biometrics matching and sample quality analysis by NIST resumed operation earlier in September after a hiatus of just over a month. The FATE Quality Specific Image Defect Detection (SIDD) evaluation assesses the effectiveness of facial image quality assessment algorithms (QAAs). It was put on hiatus from early August to September 8, along with the FRTE and IREX evaluations, to update NIST’s computing infrastructure and biometric datasets.

With the entry-to-visa dataset, Daon scored the lowest in both false non-match rate (FNMR) and efficiency among entry images after the removal of the five percent with the lowest quality scores.

Other submissions to FATE Quality this year are from Guangzhou, China-based Pixel Solutions, listed as “Pixelall” (with submissions in March and July), Mobbeel Solutions and Innovatrics.

Pixel Solutions had the lowest FNMR and the highest efficiency with the kiosk-to-entry dataset.

An algorithm submitted by Fraunhofer IGD in December had the best results by FNMR and efficiency with the entry-to-visa dataset with the bottom 1 percent of images by quality removed.

Notably, algorithms submitted in 2022 by Intema and Idemia, and in 2023 by secunet remain amongst the top five in multiple categories, underlining the uneven extent of recent gains.

“Quality assessment is critical to delivering reliable biometric capabilities,” says Michael Peirce, chief scientist at Daon. “Our algorithms identify subtle quality indicators that directly translate to better outcomes for our customers’ identity processes. By filtering out problematic images upfront, organizations achieve higher accuracy rates while reducing false rejections that frustrate legitimate users.”

Article Topics

biometric data quality  |  biometric testing  |  Daon  |  Face Analysis Technology Evaluation (FATE)  |  FATE Quality Specific Image Defect Detection (SIDD)  |  Innovatrics  |  Mobbeel  |  NIST  |  Pixel Solutions