iris recog

8/6/2019 Iris Recog.

1/36

IRIS RECOGNITION

Prepared By: 1) Vora Aarohi (7280)

2) Parekh Vandana (82916)

3) Umrigar Madhurika (82917)4) Naik Dhara (82919)

Guided By: Prof. Mita Paunwala

Electronics And Communication Dept., C.K. Pithwalla College Of

Engineering And Technology


2/36

Fingerprint Verification

Hand Geometry

Retinal Scanning

Iris Scanning

Facial Recognition

Signature Verification

Voice Verification

Types ofof biometricbiometric


3/36

Iris

Fig 1 eye image [1]


4/36

Characterist

ic

Finger

Print

Hand

Geometry

Retina Iris Face Signature Voice

Ease of use High High Low Medium Medium High High

Error

Incidence

Dryness,Di

rt,Age

Hand

Injury,Age

Glasses Poor

Lighting

Lighting,A

ge,Glasses

Changing

Signatures

Noise,Cold

s,Weather

Accuracy High High Very High Very High High High High

Cost * * * * * * *

UserAcceptance

Medium Medium Medium Medium Medium Medium High

Required

Security

High Medium High Very High Medium Medium High

LTS High Medium High High Medium Medium Medium

Why Iris ??


5/36

Criminal identification

Prison security

ATM Aviation security

Border crossing controls

Database access

Application


6/36

How Iris System works?

IMAGECAPTURE

EYELOCALISATION

IRISSEGMENTATION

NORMALISATION

FEATURE

EXTRACTION

MATCHINGDECISIONMAKING

DATABASE

ACCEPT

REJECT


7/36

ROCReceiver Operation Characteristics

EEREqual Error Rate

FARFalse Acceptance Rate

FRRFalse Rejection Rate

DOFDegree of Freedom

Performance Parameters


8/36

Segmentation

Segmentation is to isolate the actual iris region in a digital eye image.

Implementation Techniques: (1) Linear and circular Hough transform(2) Daugmans Integro-differential operator

(3) Parabolic Hough Transform


9/36

Segmentation

Input eye image Gamma adjusted image Non maximum

suppressed image

Hysterisis

thresholding image

Segmented imageSegmented iris and pupil

from original image


10/36

Gamma corrected image

The localised eye image from the database is passed through the Gaussian filter for

smoothening purpose. Calculate the gradient of the image and its orientation and perform

gamma corrections on this image for enhancement of this image.


11/36

Non maximum suppression

To get the thin fine edges of the image we perform non maxima supression. As

seen from this figure, the edges of image are not sharp in gamma corrected image

i.e. there are some reflections .So to get the thin sharp edges we perform non

maxima supression in wh

ich

we find th

e maximum of th

e pixel value around agiven centre pixel and supress the pixel which is lower than it.


12/36

Hysterisis Thresholding

To get the pixel values of the points that define the most probable edges of iris, pupil and

eyelid regions we perform hysterisis thresholding. In this, two threshold values are used to

get the soft decision. So the number of edges to be given to Hough accumulator will be

reduced.


13/36

Hough transformation

For segmenting the iris and pupil circles we calculate first the centre and radius of iris

and pupil regions.

Now using this radius and centre cordinates we overlay circles on iris and pupil region.

Also, the line is placed on the upper and lower eyelid sections and hence get thesegmented iris and pupil region.


14/36

Normalization

Normalisation transforms the iris region so that it has fixed dimensions in order

to allow comparisons.

Techniques available: (1) Daugman rubber sheet model

(2) Virtual circles


15/36

Normalization

DAUGMAN RUBBER SHEET MODEL:

Remaps eac

hpoint wit

hin t

he iris region to a pair of polar coordinates (r , )where r is on the interval [0,1] and is angle [0,2].

The iris region is modelled as a flexible rubber sheet anchored at the iris

boundary with the pupil centre as the reference point.

Fig 2 Daugmans rubber sheet model.[1]


16/36

Normalization

The remapping of the iris region from Cartesian coordinates to the normalised non-

concentric polar representation .

, , , ,I x r y r I rU U UpWith

, 1

, 1

p i

p i

x r r x rx

y r r y r y

U U U

U U U

!

!

Where, I(x,y) is the iris region image, (x,y) are the original Cartesian coordinates, (r,)

are the corresponding normalised polar coordinates, and xp,yp, and xi,yi, are the

coordinates of the pupil and iris boundaries along the direction.


17/36

For normalisation of iris regions a technique based on

Daugmans rubber sheet model is employed.

The centre of the pupil is considered as the reference

point, and radial vectors pass through the iris region, as

shown in Figure .

A number of data points are selected along each radial

line and this is defined as the radial resolution.

The number of radial lines going around the iris region

is defined as the angular resolution.

Since the pupil can be non-concentric to the iris, a

remapping formula is needed to rescale points depending

on the angle around the circle.

Normalization

Fig3. normalization [1]


18/36

Result of NormalizationResult of Normalization

Segmented imageNormalised eye

patternSegmented image

Normalised eye

pattern

Results For Normalization

Noise free normalised pattern


19/36

This is the normalized eye pattern of the segmented image obtained from the

applying remapping formula .

Normalised eye pattern

2 2

1r rEF EF Ed! s


20/36

The noisy regions inside the normalized polar array are masked with NaN

values.

Nan values are assigned by averaging of the polar array.

Noise free normalised pattern


21/36

-Feature Extraction

Gabor Filters

Log-Gabor Filters

Zero-crossings of the 1D

wavelet

HaarWavelet

Laplacian of GaussianFilters

Techniques


22/36

Feature Extraction

Log Gabor filter over the Image domain

Where,

f0 =the centre frequency.

=the bandwidth of the filter.

2

0

0

(log ( ))

( ) exp ( )

2 (log ( ))

f

fG f

fW

!


23/36

Feature Extraction

Fig 4 Feature Encoding [1]

Result of Gabor filter

Template of 40 * 480 (logical)


24/36

Feature Extraction

Haar wavelet over the Image domain

Fig 5 Haar wavelet transform [2]

if coef(i) >=0 then coef (i)=1

if coef(i)


25/36

For matching, the Hamming distance is chosen as a metric for

recognition, since bit-wise comparisons are necessary. This comparison

is done by EX-OR logic.

Matching

Hamming distanceHamming distance

1

1

( )N

j j

j

H D

N x X O R Y

!


26/36

FAR-FRR

number of samples that are falsely accepted

Total number of samples

FAR=

number of samples that are falsely re ected

total number of samples of the same eyeFRR=


27/36

HD Decision

Hamming Distance FAR FRR

0.1 0 126

0.2 0 126

0.3 0 49

0.4 0 11

0.42 1 6

0.45 53 2

0.46 110 2

0.47 260 1

0.48 431 0

0.49 518 0

0.5 530 0

0.6 531 0

0.7 531 0

0.8 531 0

0.9 531 0

1.0 531 0


28/36

FAR-FRR

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

100

200

300

400

500

600

Hamming D is tance

FAR

,FRR

FAR

FRR

Log GaborF ilter


29/36

Hamming Distance FAR FRR

0.1 0 126

0.2 0 126

0.3 0 49

0.4 0 11

0.41 1 4

0.45 60 2

0.46 110 2

0.47 260 2

0.48 431 0

0.49 520 2

0.5 530 0

0.6 530 0

0.7 531 0

0.8 531 0

0.9 531 0

1.0 531 0

HD Decision


30/36

FAR-FRR

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

10 0

20 0

30 0

40 0

50 0

60 0

Hamming D is tance

FAR,

FRR

FA R

FR R

Haar Wavelet


31/36

Database

For the matching purpose, we are using database ofCASIA version 1.

The Chinese Academy of Sciences - Institute of Automation (CASIA)

eye image database contains 756 greyscale eye images with 108 unique

eyes or classes and 7 different images of each unique eye captured in twodifferent sessions of one month difference. Three images were acquired in

first session and four in second session.

Due to specialised imaging conditions using near infra-red light,

features in the iris region are highly visible and there is good contrast

between pupil, iris and sclera regions.


32/36

Conclusion

Segmentation is the critical stage of iris recognition, since areas that are

wrongly identified as iris regions will corrupt biometric templates resulting in

very poor recognition.

Normalization is a process w

hic

hmakes encoding process very easy bymapping circular iris and pupil pattern into a rectangular sheet.

The encoding process only requires one 1D Log-Gabor filter to provide

accurate recognition while matching is the very critical stage for recognition of

an iris pattern. Haar Wavelet transform is used to decrease the template size.

Iris recognition is a highly reliable and accurate biometric technology.


33/36

Bibilography

[1] Libor Masek, Recognition of Human Iris Patterns for Biometric

Identification, presented at University of Western Australia,2003.

[2] Shin Young Lim Et Al. Efficient Iris Recognition through Improvement of

Feature Vector and Classifier published in ETRI Journal, Volume 23, No. 2,

June 2001[3] Cui, J., Wang, Y., Tan, T., Ma, L. and Sun, Z., A fast and robust iris

localization method based on texture segmentation, SPIE Defense and Security

Symposium, vol. 5404, pp. 401-408, 2004.

[4] Chinese Academy of Sciences Institute of Automation (CASIA) Iris

Database http://www.sinobiometrics.com

[5] A.Basit, M.Y.Javed and M.A.Anjum,A Robust method ofComplete IrisSegmentation, published at International Conference on Intelligent and

Advanced Systems,2007.


34/36

Bibilography

[6] John G. Daugman, "High Confidence Visual Recognition of Persons by a Test of

Statistical Independence", IEEE Trans. on Pattern Analysis and Machine Intelligence,

15(11), 1993, pp. 1148-1161.

[7] Wildes, R.P., "Iris Recognition: An Emerging Biometric Technology", Proc. of theIEEE, 85(9), 1997, pp.1348-1363.

[8] Wildes, R.P., Asmuth, J.C. et al., "A System for Automated Iris Recognition", Proc.

Of the Second IEEE Workshop on Applications ofComputer Vision, 1994, pp.121-128.

[9] Boles, W.W. and Boashash, B., "A Human Identification Technique, Proc. of the

IEEE, 85(9), 1997, pp.1348-1363.

[10]J. Daugman, High

confidence visual recognition of persons by a test of statisticalindependence, IEEE Trans. Pattern Analysis and Machine Intelligence, Vol.15, No.11,

Nov.1993, pp.1148-1161.


35/36

Bibilography

[11] R. Wildes, Iris Recognition: An Emerging Biometric Technology, Proc. of

the IEEE, vol.85, 1997. Proceedings of the Third International Conference on

Information Technology and Applications (ICITA05)

[12]W.W. Boles and B. Boashash., A Human Identification Technique UsingImages of the Iris and Wavelet Transform, IEEE Transaction on Signal Processing,

Vol.46, No. 4 ,1998.


36/36

r

iris recog

Documents