Related Experiment Videos
Integrating image quality in 2nu-SVM biometric match score fusion
Mayank Vatsa1, Richa Singh, Afzel Noore
1Lane Department of Computer Science and Electrical Engineering, West Virginia University, Morgantown, WV 26506-6109, USA. mayankv@csee.wvu.edu
International Journal of Neural Systems
|December 22, 2007
Summary
This study introduces an intelligent 2nu-support vector machine algorithm that fuses face and iris recognition scores with image quality for enhanced biometric verification. The method improves accuracy by analyzing image features and quality, outperforming existing fusion techniques.
Area of Science:
- Biometrics
- Computer Vision
- Machine Learning
Background:
- Face and iris recognition are key biometric modalities.
- Integrating multiple biometrics (multimodal biometrics) can improve recognition accuracy.
- Image quality significantly impacts biometric system performance.
Purpose of the Study:
- To propose an intelligent match score fusion algorithm for face and iris recognition.
- To enhance verification performance by integrating image quality information.
- To develop a robust fusion method using a 2nu-support vector machine.
Main Methods:
- Redundant Discrete Wavelet Transform (RDWT) for feature extraction.
- Computation of a composite image quality score (smoothness, sharpness, noise).
- Fusion of match scores and quality scores using a 2nu-support vector machine (2nu-SVM).
Main Results:
- The proposed 2nu-SVM fusion algorithm effectively integrates image quality with biometric match scores.
- Experimental validation on FERET face and CASIA iris databases demonstrated superior performance.
- The algorithm significantly outperformed existing fusion methods in verification accuracy.
Conclusions:
- The intelligent 2nu-SVM based fusion algorithm offers improved face and iris recognition performance.
- Integrating image quality is crucial for robust multimodal biometric systems.
- The proposed method provides a promising approach for advanced biometric verification.