View
43
Download
12
Category
Preview:
DESCRIPTION
IWSSIP 2009. Iris-based human verification system A research prototype. Gorazd Vrček, Peter Peer Computer Vision Laboratory Faculty of Computer and Information Science, University of Ljubljana Ljubljana, Slovenia. Chalkida, June 19 2009. Roadmap. Verification, biometry, iris? - PowerPoint PPT Presentation
Citation preview
Gorazd Vrček, Peter PeerComputer Vision Laboratory
Faculty of Computer and Information Science, University of Ljubljana Ljubljana, Slovenia
Chalkida, June 19 2009
IWSSIP 2009
Verification, biometry, iris?
System architecture
Results
Conclusion
Iris
Segmentation
Normalization
Feature extraction
Iris comparison
Input image? ROI? Problems (noise)? Segmentation goal? Start...
Getting information about the pupil: Pupil edge
Getting information about the pupil: Center
Radius
(1) indexXleft(2) Xz
(3) coarse center(4) indexYbottom(5) indexXright(6) Cz
(7) indexYup(8) Yz
Getting information about the pupil (outer edge):
Image smoothing Image illumination
Outer iris edge points detection Generating iris mask
Based on Dougman’s homogeneous rubber sheet
With the center in the center of the pupil
Gabor filter (2D Gabor wavelet)
Image convolution with it
The phase transformation used to convert the angles into iris template
Comparison of two iris bit templates Considering iris mask
Shift the bits and calculate again Use the minimal Hamming distance
The comparison within the class provides the comparison of seven images of a person among themselves
The comparison between classes provides the comparison of one iris image of a person with one of all other persons
Result: positive/negative Threshold for positive decision is set to
HD≤0.427
value 0.427 gives FAR 0%, FRR 11.584%
Research prototype → good results Comparison with ICE 2006 results (FAR=0.1%):
To improve: segmentation optimization, noise detection
To upgrade: integrate iris capturing sensor
Group FRR [%]Sagem-Iridian 2.31Cambridge 3.29Iritech 3.84
CVL 7.70
Recommended