Iris Recognition using feature optimization

Iris recognition is one of the most reliable security fields due to its distinguishing features. Segmentation of iris from the eye and varied background illumination are the two main problems which decreases the Recognition Rate (RR). Segmentation is performed by extracting circular features of iris. Illumination is tackled by use of adaptive histogram. We propose two step feature optimization to improve the recognition rate. Firstly, we improve the sharpness of image by increasing the peak pixel values and decreasing the crest pixel values based on the thresholds found by experiments. Secondly, we enhance the feature vectors prior to feature selection. This is performed by matching the ranks of the pixel values in the feature vectors. Pixel values which are matching in their ranks are enhanced and others are penalized. Rank is computed based on pixel values. This improves the feature selection and also improves classification margin. Proposed techniques were applied on CASIA database. Top recognition rate was found to be 100% where as average recognition rate is found to be 98.66% showing the proposed algorithms work satisfactorily on the dataset applied.