Performance Comparison of Feature Extraction Methods for Iris Recognition
J. Jenkin Winston, Hemanth D. Jude · Frontiers in artificial intelligence and applications · 2020
The traditional way of providing security is not enough for providing authentication over a large population. In this era of digital advancements and artificial intelligence, biometric security systems are transforming such security problems across the world. The characteristics of the iris pattern are extraordinarily unique. Hence it is extremely favored among other biometric modalities. In the iris image based biometric system, each image of the iris pattern is transformed into a set of distinct features by the process called feature extraction. It is one of the key steps in any recognition system. In this paper, statistical features are extracted from different domains like the histogram of intensity level, local binary pattern (LBP), histogram of oriented gradients (HoG), Eigenspace and moments of the iris image. A comparison of these different feature extraction methods for iris biometric system is discussed using minimum distance classifier. The experimental results show moment based feature extraction performs better than other feature extraction methods.