Non-Orthogonal Iris Segmentation

Abstract

The goal of this Trident Scholar project was to isolate the iris, the colored part of the eye, in a non-orthogonal, digital image of the human eye. A non-orthogonal image is an image where the eye is not looking directly at the camera. Iris pattern differs significantly between individuals (including identical twins) which allows for its use as an accurate biometric identifier. Both commercial and iris recognition system are becoming widespread in government and industry for logical security and access control. These iris recognition systems assume that captured iris are images are normal, or orthogonal, to the sensing devices and therefore search for circular patterns in the image. Off-angle, or non-orthogonal, images of irises cannot currently be used for identification because the iris appears elliptical; commercial algorithms cannot isolate an elliptical iris in order to start the identification process. This research expanded the functionality of iris recognition technology by developing a set of new algorithms to isolate a non- orthogonal iris in a digital image. The algorithmic approach to first isolate the pupil, the dark portion in the center of the eye. The pupil was isolated using bit-plane processing. The pupil appeared as a large homogenous region surrounded by insignificant noise, which allowed for easy definition of the pupil-iris boundary. Next, the limbic boundary (the outer edge of the iris) was determined in the cardinal directions and an ellipse was calculated that incorporated those points. After all boundaries were calculated, an "iris mask" was created to identify pixels in the image that contained the iris data, the only pixels of value for the identification of an individual. The functionality of the algorithm was tested using a database collected at the United States Naval Academy. Both orthogonal and non-orthogonal iris images were used to collect quantitative results.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
May 09, 2005
Accession Number
ADA437155

Entities

People

  • Bradford L. Bonney

Organizations

  • United States Naval Academy

Tags

Communities of Interest

  • Air Platforms
  • Biomedical
  • Materials and Manufacturing Processes

DTIC Thesaurus Topics

  • Authentication
  • Biometric Security
  • Biometrics
  • Computer Vision
  • Databases
  • Digital Images
  • Electrical Engineering
  • Graphical User Interface
  • Identification
  • Identification Systems
  • Image Processing
  • Image Recognition
  • Pattern Recognition
  • Recognition
  • Security
  • United States
  • United States Naval Academy

Readers

  • Computer Vision.