Ocular biometrics is a subflield of biometrics that is comprised of two techniques: iris scanning and retinal scanning. In each case, the uniqueness of the anatomical architecture of iris and retinal tissues provides the ability to achieve accurate user identification.
Since both are highly accurate, with reported error rates of less than one in one million, the optical and medical distinctions between these techniques should be considered when contemplating widespread acceptance and use.
The reason one can easily see another person’s iris with the naked eye, but cannot visualize another person’s retina, is largely due to the complexity of the optics needed to visualize the retina compared to the iris.
Daugman, John G., PhD, How Iris Recognition Works, IEEE Transactions on Circuits and Systems for Video Technology, Vol. 14, No. 1, January 2004, DOI 10.1109/TCSVT.2003.818350
In recent years there has been increasing interest in the ability to absolutely authenticate and/or identify individuals. User names, passwords, and tokens are all examples of what can and has been used for these purposes. More recently there has been interest in biometric identifiers that can through physical characteristics be utilized to assist in this process. Of all the possible biometric identifiers iris scans are the most reliable.
Summary and Discussion
Dr. Daugman has pioneered much of the work in this area. He is responsible for developing the complex iris recognition algorithms (mathematical formulae) that enable the use of this technique. This particular landmark paper explains these algorithms and presents results of over 9 million comparison iris images from four countries. Near-infrared wavelengths of light are used for the purpose of iris scanning. Iris feature detail is demodulated for its phase information using 2-D Gabor wavelets. Altogether 2048 phase bits are calculated for each iris. Dr. Daugman has improved upon his original algorithms to include ways of now taking into account confounding artifacts from eyelids, eyelashes, or hard contact lens edges. He includes algorithms to also allow for differences in pupil size and orientation of the eye. The uniqueness of failing the test of statistical independence forms the basis of iris recognition. He also discusses the decidability index. This concept has to do with the degree of certainty of absolute iris identification depending upon the degree of idealness of the camera, distance from the iris, and quality of illumination at the time the scan is done.
Of all the biometric identifiers that have been utilized, iris scans offer many advantages. One of these is the uniqueness of each iris. Even the same individual’s two eyes and the eyes of genetically identical twins, all have distinctly different irises. Furthermore, it appears that any given iris scan is stable over many years. It has also been demonstrated a 2 GHz processer can compare a particular iris scan with the scans of 1 million others in less than 2 seconds. At this point in time the iris scan is an extremely appealing biometric identifier. The work of Dr. John Daugman, by virtue of his iris scan algorithms, is extremely important in the development of this powerful tool.
Submitted by: William A. Carter
In contrast to retinal scanning, iris scanning is largely unaffected by the optical or refractive state of the eye. Common refractive disorders that could alter retinal scanning include high degrees of near-sightedness (myopia), far-sightedness (hyperopia), and astigmatism. Common corneal diseases that do not cause loss of tissue clarity, but do result in corneal “warping” (e.g., keratoconus), may also adversely affect retinal scanning accuracy. Consequently, unless a highly myopic, hyperopic, or astigmatic individual is wearing contact lenses, that individual will have a defocused retinal image when they take off their glasses for scanning.
On the other hand, the iris image is largely unaffected by the corneal refractive state.
Both retinal and iris images, however, will be altered dramatically by anything that results in a loss of corneal clarity, such as corneal scarring, corneal edema, and some cosmetic contact lenses.
The lens of the eye is anatomically behind the iris but is in front of the retina. Consequently, any opacification of the lens, called cataract formation, has the potential to alter the retinal scan, but a cataract will never affect an iris scan. This is not a small issue given the fact that cataracts are one of the most common eye diseases.
The iris sphincter is called the pupil. Constriction of the pupil limits the view of the retina. To avoid this, retinal scanners use a low intensity light in the infrared spectrum to achieve image capture without constricting the pupil in the process. Individuals with small pupils will give unreliable retinal scans, and it is not practical to set aside 20 to 30 minutes to pharmacologically dilate pupils when immediate identification is expected from any biometric system. Unfortunately, advancing age is one of the most common causes of papillary constriction (miosis). Bright ambient light such as outdoor daytime screening situations and certain medications also result in miosis.
Optical considerations notwithstanding, retinal scanning is far more likely to be affected by common systemic and ocular medical conditions than iris scanning.
The most common retinal diseases are vascular in nature. Some are systemic in nature such as diabetic retinopathy, hypertensive retinopathy, and vascular occlusive disease. Others are local, such as macular degeneration. All can profoundly affect the anatomy of the retinal vascular pattern, in turn, altering the previously normal vascular pattern of a person’s retinal scan. Furthermore, some common techniques used in treating these common diseases like retinal laser photocoagulation alter the retinal architecture and appearance markedly.
In past decades, cataract surgery often affected the iris appearance adversely, but newer techniques rarely cause those changes. Like the retina, the iris architecture may be altered surgically. Many glaucoma interventions (laser or surgical iridectomy) actually remove iris tissue thereby changing the scan.
Unlike the retina, common systemic or ocular diseases do not commonly affect the iris architecture.
In the biometrics literature a significant amount of time is spent comparing iris and retinal scanning techniques in terms of cost, ease of use, ease of image capture, and accuracy. By these measures, iris scanning either compares favorably or surpasses retinal scanning.
There are also important optical and medical distinctions to consider prior to widespread adoption of these techniques, and when these considerations are evaluated, iris scanning may again be superior to retinal scanning as a biometric identification technique.
Scott Eccarius, MD