Modeling the Iris
The Source for Professional Iris Cameras

Modeling the Iris as a Constitutional Indicator

The iris image represents a rich dataset for analysis and classification. As an indicator of individual traits, the iris image is recorded and a geometric model is developed by inference from the images. The geometric data representing the iris structure is sent to an interpretive model that infers physical and psychological attributes of the client based on empirically derived correlations between iris features and personal attributes. These inferred behavioral and structural traits are considered largely static with some long term progressive changes in the color (and to a much lesser extent, texture) acknowledged.

The Iris Object Model

The primary task is to obtain an efficient structural model of the iris in terms of its color and texture and associated variations. This is used for transforming a given iris image pair to a topological (spacial) model that includes a rigorous and standardized specification of geometric as well as qualitative attributes. Qualities such as chromatic and textural variations are also represented geometrically. Constellations of attributes may also be identified and extracted as higher-level features. The derived characterization of the iris in both quantitative and qualitative terms from the given images is referred to as the iris object model.

Iris Features

The features of the iris available from the frontal in vivo imaging include spacial variation of both color and texture. The texture is inferred from (and correlated to) the brightness variation. The iris geometry calls for a cylindrical coordinate system (Z,R,T) and is most conveniently specified using certain logical Symbol and Unit Conventions.

Z is the axial dimension with the reference plane (z=0) at the average pupil location in Z and units of millimeter (typically ranging -1.0 to +2.0). The increasing direction may be assumed either way but the common convention is to use increasing as inward, toward the retina of the eye. The normal iris has some amount of vaulting, or arching forward from the iris root at the periphery. The collarette is typically raised and located anterior to the pupillary border.

R (Rho) is radial offset or distance from a reference center axis (the midpoint or average of the pupil boundary is r=0) in millimeters. The pupil boundary is typically 1.5 to 2.5 mm in radius and the radius of the peripheral iris boundary typically ranges from 5 to 6 mm. The average iris is 12 mm horizontally and 11 mm vertically, with over 90% of the normal population within 1 mm of these values. The pupil center is typically superior and nasal to the midpoint of the peripheral iris boundary (pupil decentration is almost always less than 1 mm).

T (Theta) is the angular location in degrees, relative to the same reference center axis, and can have any of several sign conventions for the left and right eye. The standard mathematical convention for Theta is with t=0 corresponding to the 3:00 position and increasing counter-clockwise. For biometric purposes, an anatomically oriented coordinate system which facilitates left-right comparisons of bilaterally symmetric features is more appropriate. Therefore a standard convention for referencing iris sectors can be taken as 000-360 with 12:00=000=360 and increasing clockwise for the left eye and counter-clockwise for the right eye. Units can also be in hours:minutes (HH:MM) or radians (-pi<=T<=pi).

Geometric Model of the Iris

The study of patterns must consider that the bluer iris has more visible features due to a transparent anterior border layer than the browner iris with its relatively opaque anterior border layer. Much of the fiber separations and lacunae visible in the blue iris are actually situated posterior to the transparent anterior border layer (ABL) and would be obscured if the ABL were pigmented (as in a mixed or brown eye). (See for a photo of a blue iris before gold-coating and and for an EM image of the same specimen after gold coating).

The origin of both global and focal pigmentation is an ongoing research project at Miles Research with details available at and .

The Iris Interpretive Model

Secondary is to develop a specified interpretive model for inferring physical and psychological attributes from an identified object model.

There are two predominant interpretive iris models in common use today:

1. Iridology has traditionally been concerned with identifying structural and functional predispositions and/or disturbances in the body from the classification and identification of iris features.

2. Rayid is focused primarily on identifying psychological traits (personality constitution) from iris features.

Presumed correlation (in the Iridology and Rayid models) between health status and global amount of pigment is distinct from that between health status and local variations in pigment. The same applies to global and local texture factors.

Dynamic Aspect of Interpretive Model

The dynamic aspect of the iris model is the presumed correlation between changes in the iris and changes in health. The normal non-diseased iris exhibits only very subtle age-related changes. It is similar to the skin in that it experiences color changes (and to a lesser extent textural changes) over time as part of the normal age-related change. Such gross macroscopic changes are secondary to changes at the biochemical and ultrastructural levels.

The predominant belief in conventional physical iridology is that coloration is associated with toxicity and is regarded either as a side effect of a specific organ disturbance (e.g. the liver being perturbed may lead to the appearance of brown color patches sometimes called "liver spots" in the skin and/or on the iris) or more generally as evidence of a past deleterious influence of the environment on the body. Some interpretations will infer such a sign as reflecting ancestral conditions.

The interpretation according to the Rayid model is that the development of focal circumscribed pigment patches ("jewels") is secondary to a progressing inward/yin/contractive psychological trend which is taking place as an adaptation to the unique genetic and environmental circumstances that the individual has experienced. Some interpretations will also presume that this path is mostly predetermined by the fateful combination of ancestral and childhood factors.

Process Model of Vision

The process of vision (including image understanding) has been specified in bioengineering and artificial intelligence, and provides a useful model in designing computational processing of images of the iris for identification, analytic, or diagnostic purposes.

Visual processing is segmented in both structure and procedure (the eyeball does the preprocessing of image signals and presents reflective signals (derived information) to several lobes, most of the fibers ending/originating at the lateral geniculate nucleus). The anatomy of the optic tract is useful to consider in designing systems to process iris pictures.

Like Nature, Like Artifice: it is computationally advantageous to segment the functionality between the steps in the conscious experience of a visual impression. The process can be separated into stages, each one is an energy transduction, including... photo-electric, electro-magnetic, magneto-electric, electro-chemic, chemo-electric, and some (such as Tesla) would say electro-photic.

Visual Pre-Processing

The preprocessing of image data (Early Vision) includes both segmentation (to identify the iris portion of the image), geometric and qualitative modeling, extracting features based on computing various statistical metrics from the image data, and texture measurement such as the radial 2-D FFT on brightness to get the frequency spectrum of the fiber spacing. The derived data represents a geometric model of the anterior iris surface as inferred from the optical image. The preprocessing creates an object model from the image data, taking into account the channel limitations and bias.

Visual processing of photoelectric signals mainly includes switching the transfer direction of the signal based on content and other inputs. The signal is the energetic response of the visually functioning human cognitive response to the optical environment (input) as it is applied to the individual's visual sensor. Most of the feature extraction occurs as early preprocessing at the retinal stages, and then the feature-related information is distributed throughout the optic tract (which has a tree topology).

Visual Post-Processing

Later Visual Processing (higher level vision) serves the purpose of ascertaining the significance of object features found in the image. The late vision is more subjective as it seeks to assign semantic meaning to the individual object features as well as the overall picture.

Computational Approaches to Modeling Iris Interpretation

Early Vision may be accomplished with standardized filters for conditioning image data and more advanced processes for detecting,, classifying, and identifying features. This stage of processing extracts and encodes the intrinsic object attributes from the sampled images. Early processing may be accomplished either with specified procedures or as an open-ended learning process, using methodologies such as genetic algorithms, evolutionary methods, connectionist adaptive processes, or pure stochastic models.

Late Vision includes the assignment of meaning to visually identified object features by means of stochastic approximation. At this point the connectionist model is an efficient method to use in the design of a machine-learning cognitive process for inferring health status from identified features. This type of model can be extended to use other sources of data as input parameters.

For more information, email us at