CSE 40151 / 60151: Automated Identity Technology Using Iris Texture
Tuesday, 12:30-1:45 – DeBartolo B011
Professor Kevin W. Bowyer
(revised Nov. 17, 2010)
Can the texture pattern of
the iris be used to create a reliable, unique identifier for each person on
planet Earth? This course explores the technology of
iris biometrics in depth: the underlying conceptual theory, image acquisition,
image segmentation, texture analysis, matching of texture representations,
identity management systems, and a selection of current research issues. The
course includes readings from the literature, short writing assignments, and
practical experience with current commercial iris biometric technology.
Course-level Outcomes.
As a result of completing this course, you should: (1) Have
mastered the basic concepts and terminology related to biometrics, (2)
Understand the technical elements of the Daugman-style approach to iris
biometrics, (3) Be familiar with several current research issues in iris
biometrics, and (5) Be aware of the potential social impact of iris biometrics
in different cultural contexts.
Syllabus
of Topics
Tuesday, August 24 –
Biometrics Concepts and Terminology.
á
Slides for this lecture (3MB pdf).
á
Kevin
W. Bowyer, Karen Hollingsworth and Patrick J Flynn. Image understanding for
iris biometrics: a survey, Computer
Vision and Image Understanding 110 (2), 281-307, May 2008.
á
P.
Jonathon Phillips, Alvin F. Martin, Charles L. Wilson, and Mark A. Przybocki, An Introduction to
Evaluating Biometric Systems. IEEE
Computer 33 (2): 56-63 (2000).
Tuesday, August 31 –
Iris Image Acquistion.
á
Slides for this lecture (16 MB pdf).
á
Flom and Safir's 1987 iris
recognition patent.
á
John
Daugman's 1994 iris biometrics patent.
Tuesday, September 7
– Iris Image Segmentation.
Assignment #1 due.
This assignment looks at the growth of iris biometrics research. Prepare a histogram of the number of
iris biometrics papers published each year from 1990 through 2009. If your last name begins with A through
G, use IEEE XPlore to prepare your histogram, if your last name begins with H
through P, use Compendex, and if your last name begins with Q through Z, use
Inspec. Each of these three
databases is available through the Hesburgh library; go to the ÒResearch ToolsÓ
box in the upper right of the library web page, and look through the
alphabetical listing at the bottom of the box. Indicate the year of expiration of the Flom and Safir
patent. Estimate the growth rate
of the field in the last three years.
Tuesday, September 14
– Texture Analysis.
á
Matti PietikŠinenÕs
Scholarpedia article on Local binary patterns.
á John
Daugman paper on Gabor filters.
Tuesday, September 21
– Dr. Karen Hollingsworth on ÒFragile bitsÓ.
á
Slides for this
lecture. (3.5 MB pdf)
á
Karen
Hollingsworth, Kevin W. Bowyer and Patrick J. Flynn. The best bits in an iris code,
IEEE Transactions on Pattern Analysis and
Machine Intelligence 31 (6), 964-973, June 2009.
á
Karen
Hollingsworth, Kevin W. Bowyer and Patrick J. Flynn. Using
fragile bit coincidence to improve iris recognition, BTAS 2009.
Tuesday, September 28
– No class; please
participate in the iris / face viewing experiments.
Tuesday, October 5 –
Dr. Karen Hollingsworth on the effects of pupil dilation.
á
Slides for
this lecture. (2 MB pdf)
á
Karen
Hollingsworth, Kevin W. Bowyer and Patrick J. Flynn. Pupil dilation degrades
iris biometric performance, Computer
Vision and Image Understanding 113 (1), January 2009, 150-157.
á
NIST Interagency Report 7629
– IREX I: Performance of Iris Recognition Algorithms on Standard
Images, P. Grother, E. Tabassi, G. W. Quinn, W.
Salamon.
Tuesday, October 12 –
Iris code matching.
á
Slides for this lecture (3MB pdf).
Assignment #2 due.
This assignment looks at the accuracy of same-sensor versus cross-sensor
matching. In one image acquisition
session, acquire 50 images of one of your irises using each of two sensors. If
your last name begins with A through G, use the LG 4000 and the IrisGuard AD
100; if your last name begins with H through P, use the IrisGuard AD 100 and
the LG TD 100 handheld; and if your last name begins with Q through Z, use the
LG 4000 and the LG TD 100 handheld.
Manually screen your images to make sure that they are all of acceptable
quality. Compute the authentic
distributions (with images of one iris, you wonÕt have data for an imposter
distribution) for matches (a) between images taken using the first system, (b)
between images taken using the second system, and (c) across images from the
two systems. Determine if there is
a significant difference between the distributions in the different cases. Is there any evidence for one sensor
leading to more accurate matching than the other? Is there any evidence for cross-sensor matching being any
better or worse than same-sensor matching? Make a short ppt presentation of your results and
conclusions.
Tuesday,
October 19 – Fall break.
Tuesday, October 26 –
Effects of contact lenses.
á
Slides for this lecture (10MB pdf).
á
Sarah
Baker, Amanda Hentz, Kevin W. Bowyer and Patrick J. Flynn, Degradation
of iris recognition performance due to non-cosmetic prescription contact lenses,
Computer Vision and Image Understanding
114 (9), 1030-1044, September 2010.
Tuesday, November 2 –
Dr. Patrick Flynn on multi-biometrics and iris.
á
Slides for this lecture (9 MB pdf).
Tuesday, November 9 –
Presentations on cross-sensor matching experiments.
Tuesday, November 16
– Dr. Damon Woodard on ÒPeriocular-Based Recognition and ClassificationÓ.
Abstract. Recently,
iris and face have been well established as two of the most popular modalities
for human identification. Despite research advances inface/iris recognition,
the performance of facial and iris recognition algorithms decline significantly
when operating in non-ideal scenarios such as varying illumination, occlusions,
and expression changes. One method of addressing these challenges is through
the exploration of the use of new biometric traits which could aid in the
recognition/classification tasks. In this talk, I will present the results of
using two such traits, periocular skin texture and color, for use in biometric
identification and classification. In addition, I will provide an overview of
the current periocular biometrics research projects underway at Clemson
University's Biometric and Pattern Recognition Lab (BPRL) and discuss future
research directions.
Bio sketch. Dr.
Woodard is currently an Assistant Professor within the Human-Centered Computing
(HCC) Division in the School of Computing at Clemson University where he has
established the Biometrics and Pattern Recognition Lab (BPRL). He is also
serves as the Co-Director of Research Programs for the Center of Advanced
Studies in the Identity Sciences (CASIS). He received his Ph.D. in Computer
Science and Engineering from the University of Notre Dame in 2005, his M.E. in
Computer Science and Engineering from Penn State University in 1999, and his
B.S. in Computer Science and Computer Information Systems from Tulane
University in 1997. Prior to joining the faculty of Clemson University, Dr.
Woodard was a Director of Central Intelligence postdoctoral fellow. His
research interests include biometrics, pattern recognition, machine learning, and
image processing.
á
P. Miller, A. Rawls, D. L. Woodard, Personal Identification Using
Periocular Skin Texture, Proceedings of the 2010 ACM Symposium on
Applied Computing, Session: Applied Biometrics Track, Sierre, Switzerland, pg.
1496 – 1500, March 22-24, 2010.
á
Damon L.
Woodard, Shrinivas Pundlik, Philip Miller, Raghavender Jillela and Arun
Ross. On the Fusion of
Periocular and Iris Biometrics in Non-ideal Imagery, ICPR 2010.
á
Damon L.
Woodard, Shrinivas J. Pundlik, Jamie R. Lyle and Philip E. Miller, Periocular
Region Appearance Cues for Biometric Identification, CVPR Biometrics Workshop 2010.
á
Philip
Miller, Jamie Lyle, Shrinivas Pundlik and Damon Woodard, Performance
Evaluation of Local Appearance Based Periocular Recognition, Biometrics
Theory, Applications and Systems (BTAS 2010), Sept. 27-29, 2010..
á
Karen
Hollingsworth, Kevin W. Bowyer and Patrick J Flynn, Identifying
useful features for recognition in near-infrared periocular images, Biometrics Theory, Applications and Systems
(BTAS 2010), Sept. 27-29, 2010.
Assignment #3 due.
This assignment looks at same-session versus time-lapse iris
matching. This is related to, but
not the same as, what is normally termed template aging. Acquire 50 images of your same iris as
in Assignment #2, but in a different session at least two weeks after the
first. If your last name begins with A through G, use the LG 4000; if your last
name begins with H through P, use the IrisGuard AD 100; and if your last name
begins with Q through Z, use the LG TD 100. Compute the Òsame-sessionÓ and
Òacross-sessionÓ authentic distributions for this data. Determine if there is a significant
difference between the two distributions. Make a short ppt presentation of your
results and conclusions.
Tuesday, Nov 23 – Human
perception of iris textures.
á
Louise
Stark, Kevin W. Bowyer and Stephen Siena, Human perceptual
categorization of iris texture patterns, Biometrics Theory, Applications and Systems (BTAS), September 2010.
á
Kevin
W. Bowyer, Steve Lagree and Sam Fenker, Human versus
biometric detection of similarity in left and right irises, IEEE International Carnahan Conference on
Security Technology, October 2010.
á
Karen
Hollingsworth, Kevin W. Bowyer and Patrick J. Flynn, Similarity
of Iris Texture Between Identical Twins, IEEE Computer Society Workshop on Biometrics, June 2010.
Tuesday, November 30
– Overview of government reports on biometrics & social policy.
á
National
Academies report: Biometric
Recognition: Challenges and Opportunities (2010)
á
DHS
report: National
Strategy for Trusted Identities in Cyberspace (2010)
á
FBI:
Facial Recognition
Technology: A Survey of Policy and Implementation Issues (2009)
á
White
House: Cyberspace
Policy Review (2009)
á
NSTC:
Identity Management
Task Force Report (2008)
á
National
Science and Technology Council, Subcommittee on Biometrics report: The National
Biometrics Challenge (2006)
á
National
Academies: Strengthening Forensic Science In the United States: A Path Forward
(2009)
á
National
Academies report: IDs – Not That Easy (2002?)
á
National
Academies report: Biometric
Recognition: Challenges and Opportunities (????)
á
National
Science and Technology Council, Subcommittee on biometrics report: Privacy and Biometrics:
Building a Conceptual Foundation (2006)
Assignment #4 due. Choose (a) or (b) and write a two to three page paper. (a) Briefly compare the motivations and
challenges for using biometrics in IndiaÕs Unique ID project with those in
using biometrics in health care as in the video seen in the first class. (b)
Briefly outline the experiment that you would want to conduct in order to
determine whether iris matching is more accurate than single-fingerprint
matching.
Tuesday, December 7 –
Biometric template aging.
á
Sarah
Baker, Kevin W. Bowyer and Patrick J. Flynn, Empirical
Evidence for Correct Iris Match Score Degradation With Increased Time Lapse
Between Gallery and Probe Images, International
Conference on Biometrics, June 2009, 1170-1179.
á
Sam
Fenker and Kevin W. Bowyer, Empirical Evidence
of a Template Aging Effect In Iris Biometrics, Workshop on Applications of Computer Vision (WACV), January 2011.
Assignment #5 due. Choose (a) or (b) and write a two to three page paper. (a) What are the basic conclusions in
HollingsworthÕs paper about how pupil dilation affects iris biometrics? To what degree do the results in the
NIST IREX report reproduce HollingsworthÕs results and to what degree do they
go beyond HollingsworthÕs results?
(b) What is or should be the relationship between ocular biometrics and
iris biometrics? Should the same
illumination be used for both?
Final Exam Period:
December 15, 10:30 – 12:30.
Additional
resources for this class.
á
NIST IRis Exchange (IREX) web pages.
á NIST
Iris Challenge Evaluation (ICE) web pages.
á
Salil
Prabhakar's BCC 2009 talk on UID.
á
Salil
Prabhakar's CVPR talk on UID.
á
WSJ
article on Unique ID team.
á