[attachment=3134]

---

Presentation by:
Arun Kumar Passi Elec. Engg. (dual deg.)

---

> Introduction
> Model of a biometric system
> Why Iris
> Image Pre-processing
> Eyelash and Eyelid Removal
> Feature Extraction
> Gabor filter
> Log-Gabor filter
> Performance of the systems
> Time of Operations
> References
z
> Iris is the annular portion between the dark pupil and white sclera
> It has got a rich texture information which could be exploited for a biometric recognition system

> Its error rate is extremely low ^ Iris a permanent biometric
^ User acceptability is reasonable
> Real time biometric verification
> Less susceptible to spoofing

Claimed Identity
User Interface
Feature Extractor
One template
System Database
True / False
Verification Mode
User Interface
Feature Extractor
System Database
templates
User's Identity / User not identified
Identification Mode
> Images are generally acquired in near infra red illumination
> The distance between the eye and the camera may vary from - cm
> Iris diameter typically should be between - pixels for extracting good texture
> Careful selection of intensity level
^ Image localization
> Detecting the Pupillary circle
> Detecting outer Iris circle
^ Image normalization ^ Image enhancement
Li Ma's approach
> Project image in vertical and horizontal directions
> Minima of two projections will be a rough estimate of pupil center
> Binarize x region centered at that point using adaptive threshold
> Centroid of this binarized region is a better estimate of the pupil center
> Unwrap the Iris region onto a rectangular block of size x

Image Enhancement
Mean of every x block is calculated
Image obtained is resized to x using bi-cubic interpolation
Image Enhancement
> Interpolated image is subtracted from original normalized image
> Image is enhanced through histogram equalization
Eyelash & Eyelid removal
Edge detected image
Eyelash & Eyelid removal



//
Normalized Image
Normalized Image after noise removal

> Horizontal edge detection is used on the image
> Linear Hough transform is used to fit a line on lower and upper eyelid
> A horizontal line is then drawn intersecting the first line on the iris edge which is closest to the pupil

> Gabor filter
> Log-Gabor filter
> Laplacian of Gaussian filter
> Dyadic wavelet transform
> Mexican hat filter
¢¢¢¢¢¢¢

A Gabor filter is constructed by modulating a sine/cosine wave with a Gaussian
^ Provides the optimum simultaneous localization in both space and frequency
^ The centre frequency of the filter is specified by the frequency of the sine/cosine wave, and the bandwidth of the filter is specified by the width of the Gaussian
^ Daugman makes uses D Gabor filters in order to encode iris pattern data
> The output of Gabor filter is then demodulated to get the phase information which is quantized to four levels for each possible quadrant in complex plane

> The enhanced image is convolved with a bank of Gabor filter at different frequencies and orientations
Source: Li. Ma, Y. Wang, and T. Tan, "Iris recognition based on multi-channel Gabor filtering," in Proc. th Asian Conf. Computer Vision, // vol. I, , pp. -
) Li. Ma, Y. Wang, and T. Tan, "Iris recognition using circular symmetric filters," in Proc. th Int. Conf. Pattern Recognition., vol. II, , pp. -
) Li. Ma, Y. Wang, and T. Tan, "Iris recognition based on multi-channel Gabor filtering," in Proc. th Asian Conf. Computer Vision, vol. I, , pp. -
) Daugman, J. . How iris recognition works. IEEE Trans, CSVT , --
) Daugman, J. The importance of being random: Statistical principles of iris recognition. Pattern Recognition, vol. , num. , pp. -,
) Ma Li, Tan T., Wang Y. and Zhang, D. (): Efficient Iris Recognition by characterizing Key Local Variations, IEEE Trans. Image Processing, vol ,
no., pp. -
) A. Poursaberi, B. N. Araabi, "A Novel Iris Recognition System Using
Morphological Edge Detector and Wavelet Phase Features", ICGST
International Journal on Graphics, Vision and Image Processing,,