iriscode

iriscode is a binary feature vector produced by many iris recognition systems, most notably those based on the approach developed by John Daugman. It encodes the texture patterns of an individual’s iris into a compact digital representation that can be efficiently stored and compared. A typical iriscode is a binary string of about 2048 bits, though implementations may use shorter lengths.

The iriscode is generated from an iris image through several steps. First, the iris region is segmented and then normalized to a fixed, human-unreadable form using a rubber sheet model to account for pupil dilation and iris shape. Local texture information is then extracted, often by applying 2D Gabor filters to obtain phase information. The local phase is binarized to produce the iriscode. Because some parts of the iris may be occluded by eyelids, eyelashes, or reflections, a corresponding binary mask marks unreliable bits so they can be ignored during matching.

Matching between two iriscodes involves computing the Hamming distance, which is the fraction of mismatching bits. To compensate for rotational differences of the iris, codes are compared under a range of circular shifts, and the minimum distance is used. A system-specific threshold determines whether a distance indicates a match, balancing false accept and false reject rates.

Iriscode representations are designed to be non-invertible to protect biometric privacy, and the approach supports rapid matching suitable for large-scale authentication and identification applications. The concept has become a standard reference in iris recognition literature and commercial deployments.