Iris recognition refers to the automated method of identifying or confirming the identity of a subject by analyzing the random pattern of the iris. Iris recognition is relatively young, being only commercially developed the last decade mostly due to previous patent limitations.
The iris is a muscle within the eye, which regulates the size of the pupil and thus controlling the amount of light that enters the eye. The iris is the colored part of the eye.
Iris recognition uses the random, colored patterns within the iris. These patterns are unique for each individual.
The iris begins to form as soon as the third month of gestation, by the eighth month the structures creating the iris patterns are largely complete however pigment accretion can continue during the first postnatal years. Because the iris patters are formed randomly, even genetically identical twins will not have the same iris patterns.
The color of the iris is determined mainly by the density of melanin pigment, blue irises result from an absence of pigment.
How iris recognition works
The following sequence applies to both enrollment and recognition:
- Capture iris image. The camera acquires an image from the iris, lighting is mostly done with Near Infrared (NIR) light because with NIR there is less noise in the image due to reflections when compared to visible light. Also NIR light does not cause harm or discomfort to the subject.
- Finding iris in the image. One of the challenging parts of iris recognition is for the algorithm finding the concentric circular outside boundaries of pupil and iris. Often part of the iris is covered by eyelids or eyelashes, which even more complicates this step.
- Convert image. The set of pixels which cover the iris on the image are then transformed into a bit pattern that preserves required information for template comparisson but allows faster and statistical meaningful comparisson. Dr. Daugman’s algorithms, refered to as IrisCode ™, translates the visible characteristics from the image into a 512 byte code, the template, which allows extremely quick searches and a very low false acceptance rate.
When a subject tries to authenticate or identify himself, the generated IrisCode is compared with templates stored in the database. A test of statistical independence determines whether the IrisCode resulting from the scan and a stored IrisCode template are from the same iris.
IrisCode ™ is based on an algorithm developed and patented in the nineties by Dr. Daugman. This is nowadays the most used algorithm in commercial devices, thanks to its speed of matching with very low false match rates.
With IrisCode the visible characteristics of the iris are transformed into a phase sequence, which contains information on the orientation, spatial frequency and position of segments in the iris. The phase is not affected by contrast, camera gain or illumination levels.
The result of the algorithm is an IrisCode, 256 bytes of data which describe the phase characteristics of the iris in a polar coordinate system.
During recognition the difference between iriscodes is determined, this difference is called the Hamming Distance. If the Hamming Distance indicates that less than one third of the iriscodes is different from each other, then it is concluded that there is no statistical significant difference in between the IrisCodes and they are considered to come from the same iris.
Using 256 byte templates for all scanned irises allows matching up to 500.000 templates per second using IrisCode.
Application of iris recognition
The fact that most governments have invested in other biometric technologies such as fingerprint recognition, as well as patent limitations have led to the fact that less progress was made on commercial developments than could have been possible.
However a number of vendors already market commercial iris recognition solutions, these are mostly targeted at governments or large corporations. A few vendors are listed hereunder:
- Iris ID Systems, a spin-off of LG electronics
- IrisGuard, a company specialized in large scale security solutions based on iris recognition
Suitability of iris recognition
How suitable is iris recognition as a biometric solution? We use the following 7 criteria to evaluate the suitability of iris recognition:
|Universality||Iris recognition is said to have a very low FER (Failure to Enroll Rate), i.e. the smallest group of people which can not use the technology.|
|Uniqueness||The patterns of the iris are highly variable, and considered unique for each individual. The patterns are formed randomly during embryonic gestation, therefore even genetically identical twins have different iris patterns.|
|Permanence||The iris has the great advantage that it is internal, and thus well protected, but externally visible. Furthermore the iris does not change with ageing, one enrollment should be sufficient for a lifetime with the exception of damage due to accident or disease.|
|Collectability||The shape of the iris is almost completely flat and thus very predictable, much more than that of the face. Also an image can be taken from 10 cm up to a few meters away. Therefore no expensive 3D camera’s are needed and the impact of a different viewing angle is far less than for example with face recognition.
An issue however is the lighting, because the eyes are a reflective surface and also subjects do no want to lose their sight due to excessive lighting. Furthermore the iris is often partially obscured by eyelids, this must be compensated by the matching algoritm.
|Acceptability||Contrary to retina scans the iris can only reveal very little medical information about the subject. Furthermore subjects do not have to be in direct contact with the biometric device or camera, which is an objection that is often raised against fingerprint recognition.
A concern that some people might have is eye safety, make sure that biometric devices comply with illumination safety standards IEC 60825-1 and ANSI RP-27.1-96.
|Circumvention||A common issue with all biometric solutions is liveness detection, for non-supervised applications liveness detection is absolutely required. There are some possibilities for liveness detection on iris scanners: verification of changing pupil size upon light intensity variation, verification of the natural movement of the eyeball, using the red-eye effect, etc.|
|Performance||Iris recognition using IrisCode ™, which is used in most commercial iris recognition products, is well suited for one-to-many identification because of high speeds of comparison. Furthermore the IrisCode matching algorithm has a very low, even unprecedented false acceptance rate.|