Approved for: 2 GIAC CPEs
Iris recognition, as currently practiced, proceeds along a path: iris image capture, segmentation/normalization, demodulation/template generation, template matching, decision. Segmentation/normalization is, arguably, the most difficult of these steps. The current methods for segmentation/normalization have been successful to date. However, as we deploy large scale systems with larger populations in more challenging environments, we need to consider improvements to the segmentation/normalization process that will enable us to improve the separation between imposter and authentics distributions, particularly for cases where the imagery is not pristine. This session will focus on these challenges.
2:45-3:10
Effects of Short Term Dilation Variation on Iris Matching
George W. Quinn
3:10-3:35
The Biological Aspects of Pupil Dilation and Its Relationship in the Segmentation, Normalization, and Recognition Process
Antwan Clark
3:35-3:45
Panel Q&A
4:00-4:25
Improving Invariance to Non Circular Iris Boundaries and to Pupil Dilation
Emine Krichenn
4:25-4:50
Matching Inaccurately Normalized Iris Images
James Bergen
4:50-5:00
Panel Q&A