NIST zooms in on iris recognition
Irex program aims to help mature the biometric authentication method as a tool for PIV cards
- By William Jackson
- Nov 20, 2009
The National Institute of Standards and Technology is working with industry to encourage development of standards for iris recognition technology, and a recent testing program focused on small images that could fit on identification cards.
The Iris Exchange program, or Irex, “is about image-based interoperability,” said Patrick Grother, a supervisory computer scientist at NIST and co-author of a report on recently completed algorithm tests.
The tests were part of the first phase of Irex and included 19 algorithms from 10 providers, demonstrating strengths and weaknesses when used with various image formats and data compression methods defined by the International Organization for Standardization.
“As the largest independently administered test of iris recognition technology to date, Irex I includes a formal evaluation of the state of the art of iris recognition algorithms,” the report states. The tests showed significant improvement in recognition of small template files and resulted in recommendations for some changes in international image standards that are under consideration.
The goal of Irex is to help develop a reliable technology to be used for identity verification on distributed systems with formats such as the government’s Personal Identity Verification smart ID card. That requires a high degree of accuracy when used across a variety of hardware and software products, and compact files must be stored on devices with limited memory.
Fingerprint recognition is the most commonly used form of biometric authentication. However, no biometric verification is perfect, and a second mode of authentication is desirable, Grother said. The number of algorithms tested in the recent project illustrates the growth of interest in iris recognition. The technique is attractive because irises are as distinctive as fingerprints. But images can be captured at a distance, and the technology has a separate set of failure characteristics.
“It’s another option that may be suitable for certain environments and requirements,” Grother said. “There has been substantial vendor and government capital put into the industry in the past few years. A second modality is being bet on by the people putting money into this.”
Iris recognition is not as mature as fingerprints, Grother said. “But it’s getting there.”
Organizations whose algorithms were tested in Irex I were Cambridge University, Cogent Systems, Crossmatch Technologies, Honeywell, Iritech, L1 Identity Solutions, LG, Neurotechnology, Retica Systems, and Sagem. The algorithms were tested using three image formats and three compression methods defined in the ISO-IEC 19794-6 standard. The image format test showed that two of the three formats performed well. Two of the compression standards were also found to perform well enough for possible use on smart cards.
“Standard iris image records with size of approximately 30 kilobytes can be produced for large-scale identification applications,” the study concluded. “This represents a factor of 10 reduction in size over the images captured using contemporary cameras.”
For one-to-one image comparisons, “standard iris image records with sizes around 3 kilobytes can be produced that are suitable. This factor of 100 reduction in size over the images captured using contemporary cameras makes the images suitable for storage on ‘smart card’ credentials.”
Grother said recommendations for changes to the standard included how the format crops the iris image being compared. “They are not tolerant of too much cropping,” he said of the algorithms. He said the standard is under revision and should be completed in 2010.
NIST is expected to complete guidelines based on the tests for agencies using iris recognition by the end of the year, he said. A second round of Irex testing of standard images is scheduled to take place during the next nine to 12 months. Results from the Irex I tests are available online at iris.nist.gov/irex. A draft of Irex II research plans is available at iris.nist.gov/irexII.