DSP-Based Implementation and Optimization of an Iris Verification Algorithm using Textural Feature
Author: Richard Ng Yew Fatt Co-authors: Mr. Mok Kai Ming Dr. Tay Yong Haur
Computer Vision and Intelligent Systems (CVIS) Group Universiti Tunku Abdul Rahman, Malaysia Presented by: Tou Jing Yi
1. INTRODUCTION - What is biometrics?
It employs physiological or behavioural characteristics to identify an individual.
Physiological characteristics Iris, fingerprint, face, hand geometry and palm-prints.
Behavioural characteristics Voice, signature and keystroke dynamics.
1.1 Iris Verification Stages
2. PROPOSED ALGORITHM 2.1 Iris Segmentation Circular Hough transform is used to detect iris inner boundary [4]. Edge detector is applied to generate the edge map.
Gaussian filter is applied to smooth the image to select the proper scale of edge analysis. The center coordinate and radius of the circle with maximum number of edge points are defined as the pupil center and iris inner boundary.
Fig . Iris inner boundary localization.
( xi xc ) 2 ( yi yc ) 2 r 2
2.1 Iris Segmentation The proposed method selects two search regions including the outer iris boundaries. The intensities of each radius in the two search regions are added up. The right and left iris boundaries are the maximum difference between the sum of intensities of two outer radii and two inner radii. Fig . Iris outer boundary localization.
2.2 Iris Normalization Iris may be captured in different size with varying imaging distance.
Normalization process
Fig . Original image
Radial resolution
Normalize the iris region into rectangular block.
Angular resolution Fig . Normalized image
2.3 Iris Enhancement
Normalized image
Enhancement
Enhanced image
• The normalized iris image has low contrast and non-uniform illumination caused by the light source position
• Local histogram equalization is applied to reduce the effect of non-uniform illumination and obtain well-distributed texture image • Enhanced image is divided into three zones according to the characteristic of the iris texture
2.4 Feature Extraction exp(( log( w / w0 ) 2 ) G ( w) 2(log( k / w0 )) 2 ) where w0 = Filter’s center frequency k = Bandwidth of the filter 1D Log Gabor filter is used to extract the frequency information which represents the iris textures [5]. Zone Z1, Z2 and Z3 are processed with filter of decreasing center frequency respectively.
2.4 Feature Extraction Each pattern is demodulated to extract its phase information.
The phase information is quantized into four quadrants in the complex plane.
Fig. Phase-quadrant demodulation code [2-3]
2.5 Template Matching Hamming distance is a measure of dissimilarity between two binary templates. A threshold is set to decide if the two templates are from the same person or different persons. HD
(templateA templateB ) maskA maskB maskA maskB
THD HD1 HD2 HD3
1
α, β and γ have decreasing weightings because inner zone Z1 provides more texture information than outer zone Z3.
3. PERFORMANCE EVALUATION We run experiments on CASIA iris image database version 1.0 [1]. There are 756 iris images from 108 different irises. For each eye, 7 images are captured in two sessions. The time interval between two sessions is about one month. The resolution of the iris images is 320×280 pixels.
Fig. Localized iris images
3. PERFORMANCE EVALUATION Fig. ROC Curve for iris verification system
4. DEVELOPMENT ENVIRONMENT
IDDE – Integrated software development and debugging environment. It includes C/C++ compiler, assembler, expert linker, loader, run-time library, simulator and hardware emulator. The embedded iris verification system is implemented on ADSPBF561 EZ-KIT LITE evaluation board. It consists of decoding part, encoding part and DSP processing part.
4.1. System architecture for iris verification system
4.2 Optimization and Performance Profile
The optimization strategies such as source code tuning, compiler’s pragmas, and conditional code optimization are implemented.
Data cache and memory optimization are implemented to utilize the memory hierarchy.
The performance is evaluated in term of speed before and after optimization.
4.2 Optimization and Performance Profile
The performance profile for the iris verification algorithm is gathered using statistical profiler tool.
After optimization, the total computation time of iris verification algorithm drops for about 67%.
The total computation time is 0.475 second, which conforms to the speed requirement of the iris verification system.
4.2 Optimization and Performance Profile Total computation time for IVS before and after optimization
5. CONCLUSIONS
A robust iris verification algorithm is implemented on Blackfin DSP. The algorithm has achieved a high recognition rate of 98.32%. The iris verification algorithm is mapped and optimized on the EZ-KIT Lite evaluation board. The total verification time is 475.23 milliseconds to process an iris image. DSP-based iris verification system is portable, power efficient, fast authentication and compact in size.
6. REFERENCES [1] “CASIA iris image database,” http://www.sinobiometrics.com/Databases.htm, 2007. [2] J. Daugman, “High confidence visual recognition of persons by a test of statistical independence,” IEEE Tans. Pattern Analysis and Machine Intelligence, vol.15, 1993, pp.11481161. [3] J. Daugman, “How iris recognition works,” IEEE Trans. CSVT, vol. 14, no. 1, 2004, pp. 21 – 30. [4] R. Y. F. Ng, Y. H. Tay, and K. M. Mok, “An effective segmentation method for iris recognition system,” Proc. the Fifth International Conference on Visual Information Engineering, July 2008, pp. 548-553. [5] R. Y. F. Ng, Y. H. Tay, and K. M. Mok, “Iris Recognition Algorithms Based on Texture Analysis,” Proc. 3rd International Symposium on Information Technology, vol. 2, Aug 2008, pp. 904-908.
7. Questions?
Q&A Richard (
[email protected])
