A Robust Human Iris Verification Using a Novel Combination of Features

This paper proposes a method for personal identification based on iris recognition. The iris segmentation is obtained by using an integro-differential operation. The segmented iris is then normalised and actually a small portion of the normalised portion is used for feature extraction. Three types of features are used: GLCM based features, Edge based features and Local Directional Pattern. We present a comparison of the performances of the above mentioned features. It is also shown experimentally that the half-way iris patterns exhibit a symmetry about the vertical axis. The multiclass recognition problem is reduced to a two class verification problem. Experimental results show that our proposed method has encouraging performance.