Computer vision & A.I.

VeriEye Algorithm Features and Capabilities

Download VeriEye SDK brochure (PDF)

Neurotechnology began research and development in the field of eye iris biometrics in 1994. In 2008, Neurotechnology released VeriEye iris recognition algorithm. The next year VeriEye was recognized by NIST as one of the most reliably accurate iris recognition algorithms.

The proprietary algorithm implements advanced iris segmentation, enrollment and matching using robust digital image processing algorithms:

  • Robust iris detection. Irises are detected even when there are obstructions to the image, visual noise and/or different levels of illumination. Lighting reflections, eyelids and eyelashes obstructions are eliminated. Images with narrowed eyelids or eyes that are gazing away are also accepted.
  • Automatic interlacing detection and correction results in maximum quality of iris features templates from moving iris images.
  • Gazing-away eyes are correctly detected on images, segmented and transformed as if it were looking directly into the camera (see Figure 1).
  • Correct iris segmentation is obtained even under these conditions:
    • Perfect circles fail. VeriEye uses active shape models that more precisely model the contours of the eye, as iris boundaries are not modeled by perfect circles.
    • The centers of the iris inner and outer boundaries are different (see Figure 2). The iris inner boundary and its center are marked in red, the iris outer boundary and its center are marked in green.
    • Iris boundaries are definitely not circles and even not ellipses (see Figure 3) and especially in gazing-away iris images.
    • Iris boundaries seem to be perfect circles. The recognition quality can still be improved if boundaries are found more precisely (see Figure 4). Note these slight imperfections when compared to perfect circular white contours.
  • Reliability. VeriEye 2.9 algorithm shows excellent performance when tested on all publicly available datasets. Especially good results are achieved on the recent NIST ICE2005 Exp1 database with iris images of intentionally degraded quality. See testing results for more details.

All iris images are taken from CASIA Iris Image Database V2.0 and CASIA Iris Image Database V3.0 collected by the Chinese Academy of Sciences Institute of Automation (CASIA) (

Technical Specifications

64 pixels is the minimal radius of circle containing full iris texture, that is required for iris template extraction.

Near-infrared spectral region is recommended for iris image capture.

All iris templates should be loaded into RAM before identification, thus the maximum iris template database size is limited by the amount of available RAM.

VeriEye biometric template extraction and matching algorithm is designed to run on multi-core processors allowing to reach maximum possible performance on the used hardware.

VeriEye 2.9 iris engine specifications
  Android-based (1)
PC-based (2)
Template extraction components Embedded
Template extraction time (seconds) 1.34 1.20 1.34 0.60
Template matching components Embedded Iris Matcher Iris Matcher
Template matching speed
(irises per second)
3,000 40,000
Single iris record size in a template (bytes) 2,328

(1) Requires to be run on Android devices based on at least Snapdragon S4 system-on-chip with Krait 300 processor (4 cores, 1.51 GHz).
(2) Requires to be run on PC or laptop with at least Intel Core 2 Q9400 quad-core processor (2.67 GHz) to reach the specified performance.

Reliability Test Results

We present the testing results to show VeriEye 2.9 template matching algorithm reliability evaluations. Iris images from several standard databases were used for testing, thus the testing results can be compared with testing results of other algorithms. All databases contained iris images with 640 x 480 pixels size.

Iris image databases used for VeriEye 2.9 algorithm testing
Database name Images quantity Persons quantity Unique eye quantity
ICE2005 Exp1 iris image database (Right Iris) 1,425 124 124
University of Notre Dame, ND-IRIS-0405 64,980 356 712
University of Bath, IRISDB1600 (1) 24,361 624 1231

(1) The full IRISDB1600 database contains 31,997 images (image size 1280x960 pixels), representing 799 unique persons and 1,598 unique irises. A subset used in this test was preprocessed similar to NIST IREX experiments:
    (a) Images were downsampled to 640x480 via 2x2 neighborhood averaging.
    (b) All images containing irises with diameters larger than 340 pixels were removed.

Two tests were performed with each database:

  • Test 1 maximized matching accuracy. VeriEye 2.9 algorithm reliability in this test is shown as blue curves on the ROC charts.
  • Test 2 maximized matching speed. VeriEye 2.9 algorithm reliability in this test is shown as red curves on the ROC charts.

The iris rotation tolerance was set to ±15° in all tests.

Receiver operation characteristic (ROC) curves are usually used to demonstrate the recognition quality of an algorithm. ROC curves show the dependence of false rejection rate (FRR) on the false acceptance rate (FAR).

ICE2005 Exp1
VeriEye ROC charts on ICE2005 Exp1 iris image database
Click to zoom
VeriEye ROC charts on ND-IRIS-0405 iris image database
Click to zoom
Bath IRISDB1600
VeriEye ROC charts on Bath IRISDB1600 iris image database
Click to zoom
VeriEye 2.9 algorithm reliability testing results, FRR at 0.001 % FAR
  Test 1 Test 2
ICE2005 Exp1 database 0.0983 % 0.1187 %
ND-IRIS-0405 database 1.5550 % 1.6000 %
BATH IRISDB1600 database 0.0893 % 0.0928 %
AFIS or multi-biometric fingerprint, iris, face and voice identification for large-scale systems.

Face identification for PC, mobile and Web solutions.

Fingerprint identification for PC, mobile and Web solutions.

Iris identification for PC, mobile and Web solutions.

Speaker recognition for PC, mobile and Web applications.

Ready-to-use robotics development kit.

More products for developers:

End-user products:
  • NCheck Bio Attendance – an attendance control application that uses fingerprint or face biometrics to perform persons identification.
  • NVeiler Video Filter – a plug-in for VirtualDub that automatically detects faces in a frame, tracks the faces (or other objects) in subsequent frames and hides them.
  • NPointer – a freeware application for gesture- and voice-based computer control that captures hand movements with a webcam and accepts voice commands.
Copyright © 1998 - 2015 Neurotechnology | Terms & Conditions | Privacy Policy