Abstract
The central issue in pattern recognition is the relation between within-class variability and etween-class variability. These are determined by the various degrees-of-freedom spanned by the patterns themselves, and ny the selectivity of the chosen feature encoders. An interesting application of 2Dwavelets in computer vision is the automatic recognition of personal identity by encoding and matching the complex patterns visible at a distance in each eye’s iris. Because the iris is a protected, internal, organ whose random texture is highly unique and stable over life, it can serve as a kind of living password or passport that one need not remember but is always in one’s possession. I will describe wavelet demodulation methods that I have developed for this problem over the past 10 years, and which are now installed in all existing commercial systems for iris recognition. The principle that underlies iris recognition is the failure of a test of statistical independence performed on the phase angle sequences of iris patterns. Quadrature 2DGabor wavelets spanning 3 octaves in scale enable the complex-valued assignment of local phasor coordinates to iris patterns. The combinatorial complexity of these phase sequences spans a out 244 independent degrees-of-freedom, and generates binomial distributions for the Hamming Distances (a similarity metric)etween different irises. In six public independent field trials conducted so far using these algorithms, involving several millions of iris comparisons, there has never been a single false match recorded. The time required to locate and to encode an iris into quantized wavelet phase sequences is 1 second. Then database searches are performed at a rate of 100,000 irises/second. Data will be presented in this talk from 2.3 million IrisCode comparisons. This wavelet application could be used in a wide range of settings in which persons’ identities must be established or confirmed by large scale database search, without relying upon cards, keys, documents, secrets, passwords or PINs.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Author information
Authors and Affiliations
Consortia
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2001 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Daugman, J., OBE. (2001). Personal Identification in Real-Time by Wavelet Analysis of Iris Patterns. In: Tang, Y.Y., Yuen, P.C., Li, Ch., Wickerhauser, V. (eds) Wavelet Analysis and Its Applications. WAA 2001. Lecture Notes in Computer Science, vol 2251. Springer, Berlin, Heidelberg. https://doi.org/10.1007/3-540-45333-4_1
Download citation
DOI: https://doi.org/10.1007/3-540-45333-4_1
Published:
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-43034-6
Online ISBN: 978-3-540-45333-8
eBook Packages: Springer Book Archive