文档介绍：Image and puting 28 (2010) 213–214
Contents lists available at ScienceDirect
Image and puting
journal homepage: ate/imavis
Editorial
Introduction to the Special Issue on the Segmentation of Visible Wavelength Iris
Images Captured At-a-distance and On-the-move
1. Overview and are noise-free should appear as black, whereas the remaining
pixels should be represented by white (Fig. 1).
Deployed iris recognition systems are mainly based on Daug- Later, a disjoint test set of 500 images was used to measure the
man’s pioneering approach, and have proven their effectiveness pixel-by-pixel agreement between the binary maps made by each
in relatively constrained scenarios: operating in the near infra- participant and the ground-truth data, manually built by a-
red spectrum (NIR, 700-900 nm), at close acquisition distances nizers of the contest and that — a priori — should be accepted as the
and with stop-and-stare interfaces. However, the human iris optimal segmentation. The test error rates for each submission
supports contactless data acquisition, and it can — at least theoret- were the average error rates for each test image. The error rates
ically — be imaged covertly. The feasibility of covert iris recogni- on each test image were the proportion of correctly classified pix-
tion receives increasing attention and is of particular interest for els. The contest received a total of 97 participants from over 22
forensic and security purposes. In this scope, one possibility is countries. The best 8 participants (those that achieved the lowest
the use of visible wavelength light (VW) to perform image acquisi- test error rates) were invited to publish their approach in this spe-
tion, although the use of this type of light can severely degrade the cial issue, that we hope will constitute a valuable source of infor-
quality of the captured data. This is mainly due to the optical prop- mation for many of the researchers concerned with the
erties of the two