Download - Week10 Iris
-
7/27/2019 Week10 Iris
1/46
Iris Modeling and Synthesis
CPSC 601 Biometric Technologies Course
-
7/27/2019 Week10 Iris
2/46
Lecture Plan
Motivation
Iris structure
Iris Image acquisition
Methodology Iris Localization
Iris features Matching
Iris Synthesis
Future Developments
-
7/27/2019 Week10 Iris
3/46
Anatomy of the human iris. The upper panel illustrates the structure ofthe iris seen in a transverse section. The lower panel illustrates thestructure of the iris seen in a frontal sector.
Iris Structure
-
7/27/2019 Week10 Iris
4/46
At a finer grain of analysis, the iris is composed of several
layers. The posterior surface is composed of heavily
pigmented epithelial cells that make it impenetrable to light.
Anterior to this layer two muscles are located that work in
cooperation to control the size of the pupil.
The visual appearance of the iris is a direct result of itsmultilayered structure. Iris color results from the differential
absorption of light impinging on the pigmented cells in the
anterior border layer.
Iris Structure
-
7/27/2019 Week10 Iris
5/46
The first source of evidence comes
from clinical observations. Duringthe course of examining largenumber of eyes, ophthalmologistshave noted that the detailedspatial pattern of an Iris seems tobe unique. The pattern seem tovary little, at least past childhood.
The second source of evidencecomes from developmentalbiology. While the generalstructure of the iris is geneticallydetermined, the particulars of itsminutiae are critically dependenton circumstances (e.g. the initialconditions in the embryonic
precursor to the iris).
Anatomy of the iris visible in an optical
image.
Iris Structure - Uniqueness
-
7/27/2019 Week10 Iris
6/46
Another interesting aspect of the physical characteristics of the iris
from a biometric point of view has to do with its dynamics. Due to the
complex interplay of the iris's muscles, the diameter of the pupil is in a
constant state of small oscillation at a rate of approximately 0.5 Hz.
This movement could be monitored to ensure that a live specimen is
being evaluated.
Since the iris reacts very quickly to changes in impinging illumination,
monitoring the reaction to a controlled illuminant could provide similar
evidence.
Iris Structure - dynamics
-
7/27/2019 Week10 Iris
7/46
Acquisition of a high-quality iris image,
while remaining non-invasive to human
subjects, is one of the major challengesof automated iris recognition.
Figure: Passive sensing approaches toiris image acquisition. The upperdiagram shows a schematic diagram
of the Daugman image acquisition rig.The lower diagram shows a schematicdiagram of the Wildes et al. imageacquisition.
Iris Image Acquisition
-
7/27/2019 Week10 Iris
8/46
In order to cope with the inherent variability of ambientillumination, extant approaches to iris image sensing providea controlled source of illumination as a part of their method.
Research initiated at Sarnoff Corporation and subsequentlytransferred to Sensar Incorporated for refinement andCommercialization has yielded the most non-invasive approach toiris image capture that has been documented to date.
For capture, a subject merely needs to stand still and face forwardwith their head in an acquisition volume of 600vertical by
450 horizontal and a distance of approximately 0.38 to 0.76 m, allmeasured from the front-center of the acquisition rig. Capture of an imagethat has proven suitable to drive iris recognition algorithm can then beachieved totally automatically, typically within 2-10 seconds.
Iris Image Acquisition
-
7/27/2019 Week10 Iris
9/46
Figure: Active sensing approach to iris image acquisition.
Iris Image Acquisition
-
7/27/2019 Week10 Iris
10/46
Following image acquisition, the portion ofthe image that corresponds to the iris
needs to be localized from itssurroundings.
The iris image data can then be brought
under a representation to yield an irissignature for matching against similarlyacquired, localized and represented irises.
Iris Image Localization
-
7/27/2019 Week10 Iris
11/46
Daugman and Wildes et al. approaches make use of firstderivatives of image intensity to signal the location of edges
that correspond to the borders of the iris. Here, the notionis that the magnitude of the derivative across an imagedborder will show a local maximum due to the local changeof image intensity.
Both systems model the various boundaries that delimit theiris with simple geometric models. For example, they bothmodel the limbus and pupil with circular contours.
The Wildes et al. system also explicitly models the upper andlower eyelids with parabolic arcs. In initial implementation,the Daugman system simply excluded the upper and lowermost portions of the image where eyelid occlusion wasmost likely to occur; subsequent refinements includeexplicit eyelid localization.
Iris Image Localization
-
7/27/2019 Week10 Iris
12/46
The two approaches differ mostly in the way that they search their
parameter spaces to fit the contour models to the image information. The
Daugman approach fits the circular contours via gradient ascent on the
parameters (xc, y
c, r) so as to maximize
where
is a radial Gaussian with center r0and standard deviation thatsmoothes the image to select the spatial scale of edges under
consideration, *symbolizes the convolution, dsis an element of
circular arc and division by 2rserves to normalize the integral.
Iris Image Localization
-
7/27/2019 Week10 Iris
13/46
The Wildes et al. approach performs its contour fitting in two steps.First, the image intensity information is converted into a binary edge-map. Second, the edge points vote to instantiate particular contourparameter values.
Iris Image Localization
-
7/27/2019 Week10 Iris
14/46
Both approaches to localizing the
iris have proven to be successful in
the targeted application. The
histogram-based approach tomodel fitting should avoid problems
with local minima that the active
contour model's gradient descent
procedure might experience.
However, by operating moredirectly with the image derivatives,
the active contour approach
avoids the inevitable thresholding
involved in generating a binary
Edge map.
Illustrative results of iris localization.Given an acquired image, it isnecessary to separate the iris from thesurroundings. Taking as input an irisimage, automated processing
delineates that portion whichcorresponds to the iris.
Iris Image Localization
-
7/27/2019 Week10 Iris
15/46
The distinctive spatial characteristics of the human iris are displayed at a variety
of scales.
The Daugman approach makes use of a decomposition derived from
application of a two-dimensional version of Gabor filters to the image data.
Since the Daugman system converts to polar coordinates, (r, ), during
matching, it is convenient to give the filters in a corresponding form as
where and co-vary in inverse proportion to generate a set ofquadrature pair frequency selective filters, with center locations specified by (r0,
0). These filters are particularly notable for their ability to achieve good joint
localization in the spatial and frequency domains.
Iris modeling- methodology
-
7/27/2019 Week10 Iris
16/46
The Wildes et al. approach makes use of an isotropic bandpassdecomposition derived from application of Laplacian of Gaussian (LoG)filters to the image data. The LoG filters can be specified via the form
with the standard deviation of the Gaussian and the radialdistance of a point from the filters center. In practice, the filteredimage is realized as a Laplacian pyramid.
Iris modeling- methodology
-
7/27/2019 Week10 Iris
17/46
By retaining only the sign of the Gabor filter output, the
Representational approach that is used by Daugman yields a
remarkably parsimonious representation of an iris. Indeed, arepresentation with a size of 256 bytes can be
accommodated on the magnetic stripe affixed to the back
of standard credit/debit cards. In contrast, the Wildes et
al. representation is derived directly from the filteredimage for size on the order of the number of bytes in the
iris region of the originally captured image.
Iris modeling- methodology
-
7/27/2019 Week10 Iris
18/46
Iris matching can be understood as a three-stage process.
The first stage is concerned with establishing a spatialcorrespondence between two iris signatures that are to be
compared.
Given correspondence, the second stage is concernedwith quantifying the goodness of match between two irissignatures.
The third stage is concerned with making a decisionabout whether or not two signatures derive from the same physicaliris, based on the goodness of match.
Iris matching
-
7/27/2019 Week10 Iris
19/46
Given the combination of required subject participation and thecapabilities of sensor platforms currently in use, the keygeometric degrees of freedom that must be compensated for inthe underlying iris data are shift, scaling and rotation.Shift accounts for offsets of the eye in the plane parallel to thecamera's sensor array. Scale accounts for offsets along thecamera's optical axis. Rotation accounts for deviation in angularposition about the optical axis. Another degree of freedom ofpotential interest is that of pupil dilation.
Iris matching
-
7/27/2019 Week10 Iris
20/46
Daugmans system uses radial scaling to compensate for overall size
as well as a simple model of pupil variation based on linear stretching.
The scaling serves to map Cartesian image coordinates (x, y) to polar
image coordinates (r, ) according to
where r lies on [0, 1] and is cyclic over [0, 2], while (xp(), yp())
and (x1(), y1()) are the coordinates of the pupillary and limbicboundaries in the direction . Rotation is compensated for by brute
force search: explicitly shifting an iris signature in by various
amounts during matching.
Iris matching
-
7/27/2019 Week10 Iris
21/46
The Wildes et al. approach uses an image registration technique to
compensate for both scaling and rotation. This approach geometrically projects
an image, Ia(x, y), into alignment with a comparison image, Ic(x, y), according
to a mapping function (u(x, y), v(x, y)) such that, for all (x, y), the image
intensity value at (x, y)(u(x, y), v(x, y)) in Iais close to that at (x, y) in Ic.
More precisely, the mapping function (u, v) is taken to minimize
while being constrained to capture a similarity transformation of image
coordinates (x, y) to (x, y), i.e.
with sa scaling factor and R()a matrix representing rotation by .
Iris matching
-
7/27/2019 Week10 Iris
22/46
An appropriate match metric can be based on direct point wise
Comparisons between primitives in the corresponding signature
representations. The Daugman approach quantifies this matter by
computing the percentage of mismatched bits between a pair of irisrepresentations, i.e. the normalized Hamming distance. Letting A and B be
two iris signatures to be compared, this quantity can be calculated as
With subscriptj indexing bit position anddenoting the exclusive-OR operator. The Wildes et al. system employs a
somewhat more elaborate procedure to quantify the goodness of match.
The approach is based on normalized correlation between two signatures
(i.e. pyramid representations) of interest.
Iris matching
-
7/27/2019 Week10 Iris
23/46
The final subtask of matching is to evaluate the goodness ofmatch values to make a final judgement as to whether twosignatures under consideration do (authentic) or do not(impostor) derive from the same physical iris. In theDaugman approach, this amounts to choosing a separationpoint in the space of (normalized) Hamming distancesbetween the iris signatures. Distances smaller than theseparation point will be taken as indicative of authentics;those larger will be taken as indicative of impostors.
In the Wildes et al. approach, the decision making processmust combine the four goodness of match measurementsthat are calculated by the previous stage of processing (i.e.one for each pass band in the Laplacian pyramidrepresentation that comprises a signature) into a single
accept/reject judgement.
Iris matching
-
7/27/2019 Week10 Iris
24/46
Further developments could befocused on yielding more compactSystems that can be easilyincorporated into consumer productswhere access control is desired (e.g.automobiles, personal computers,various handheld devices).
Can iris recognition be performed atgreater subject to sensor distanceswhile remaining unobtrusive?
How much subject motion can betolerated during image capture?
Can performance be made morerobust to uncontrolled ambientillumination?
Toward iris recognition at a distance.An interesting direction for futureresearch in iris recognition is to relaxconstraints observed by extantsystems. As a step in this direction, aniris image captured at 10m subject to
sensor distance is shown.
Iris modelingfuture developments
-
7/27/2019 Week10 Iris
25/46
At a more operational level of performance analysis, studies ofiris recognition systems need to be performed whereindetails of acquisition are systematically manipulated,documented and reported. Parameters of interest include,geometric and photometric aspects of the experimentalstage, length of time monitored and temporal lag betweentemplate construction and recognition attempt. Similarly,details of captured irises and relevant personal accessoriesneed to be properly documented in these same studies
(e.g. eye color, eyewear).
Iris modelingfuture developments
-
7/27/2019 Week10 Iris
26/46
Iris Synthesis
Classical Biometrics - Recognition Fingerprints, Faces, Irises
Inverse Problem Synthesis
Testing Recognition methods
-
7/27/2019 Week10 Iris
27/46
Iris Synthesis - Goals
Synthesis Of Biometric Databases
Iris Database Augmentation
Testing Recognition Methods
Minimal User Input
-
7/27/2019 Week10 Iris
28/46
Iris Synthesis - previous work
Iris Recognition - [Wildes 94, Daugman 04]
Biometric Synthesis - [Yanushkevich et al.04]
Iris Synthesis - [Lefohn et al. 03, Cui et al. 04]
-
7/27/2019 Week10 Iris
29/46
Iris Synthesis
An Ocularists Approach to Human
Iris Synthesis. [ Lefohn et. al. 03]
An Iris image synthesis method
based on PCA and Super-Resolution. [Cui et. al. 04]
-
7/27/2019 Week10 Iris
30/46
PCA Approach
Uses 75 Dimensional PCA Feature
Vector Randomization
Super Resolution
Great statistical results Low Realism
-
7/27/2019 Week10 Iris
31/46
Ocularists Approach
Uses: 30-70
Layers Great Results.
Domain Specific
KnowledgeAn ocularist's approach to human iris synthesis. Lefohn et. al. 2003.
Used with permission.
-
7/27/2019 Week10 Iris
32/46
New Approach
Use Real Iris Sample Use sets of Similar Irises
Capture Characteristics Chaikin Reverse Subdivision
Combine Characteristics Multiple Iris Donors
See L. Wecker, F. Samavati and M. Gavrilova workson the subject.
-
7/27/2019 Week10 Iris
33/46
Comparison
PCA Global Features
Not as Efficient
Realism
Reverse
Subdivision Global & Local Features
Linear Implementation
Realistic results
-
7/27/2019 Week10 Iris
34/46
Organization
-
7/27/2019 Week10 Iris
35/46
Method
First step: Isolate the iris.
Polar Transform
Iris Stretching
-
7/27/2019 Week10 Iris
36/46
Multiresolution
Data has many resolutions Levels of resolution have different meanings
Reverse Subdivision Details
-
7/27/2019 Week10 Iris
37/46
Decomposition
-
7/27/2019 Week10 Iris
38/46
Method
Capture Details Reverse Subdivision
Details All Characteristics
Courtesy of: Michal Dobes and Libor Machala,
Iris Database, http://www.inf.upol.cz/iris/
http://www.inf.upol.cz/iris/http://www.inf.upol.cz/iris/http://www.inf.upol.cz/iris/ -
7/27/2019 Week10 Iris
39/46
Combinations
-
7/27/2019 Week10 Iris
40/46
Classifications
Frequency of Data
Number of Concentric Rings
Courtesy of: Michal Dobes and Libor Machala, Iris Database, http://www.inf.upol.cz/iris/
http://www.inf.upol.cz/iris/http://www.inf.upol.cz/iris/http://www.inf.upol.cz/iris/ -
7/27/2019 Week10 Iris
41/46
Database Size
-
7/27/2019 Week10 Iris
42/46
Courtesy of: Michal Dobes and Libor Machala, Iris Database, http://www.inf.upol.cz/iris/
Original SetInput Irises
http://www.inf.upol.cz/iris/http://www.inf.upol.cz/iris/http://www.inf.upol.cz/iris/ -
7/27/2019 Week10 Iris
43/46
Courtesy of: Michal Dobes and Libor Machala, Iris Database, http://www.inf.upol.cz/iris/
Output Irises
http://www.inf.upol.cz/iris/http://www.inf.upol.cz/iris/http://www.inf.upol.cz/iris/ -
7/27/2019 Week10 Iris
44/46
Combinations
-
7/27/2019 Week10 Iris
45/46
Output Irises
Courtesy of: Michal Dobes and Libor Machala, Iris Database, http://www.inf.upol.cz/iris/
http://www.inf.upol.cz/iris/http://www.inf.upol.cz/iris/http://www.inf.upol.cz/iris/ -
7/27/2019 Week10 Iris
46/46
Future Work
Post-Processing
Multiple samples of each iris
Verification
Statistically