DFT-DCT Combination Based Novel Feature Extraction Method for Enhanced Iris Recognition

Iris Recognition (IR) using conventional methods is a challenging domain, and incorporating a combination of two transforms along with proposed novel extraction technique possesses the efficacy to address the problem at hand. This paper throws light upon the proposed unique Combined DFT-DCT feature extraction along with the inclusion of a disc shaped morphological structuring element in the preprocessing stage. Two novel methods, namely astroid and astroid ring shaped extraction techniques are proposed, and Binary Particle Swarm Optimization (BPSO) based algorithm for feature selection has been employed to procure the optimal subset of features from the feature space. Experimental results that have been obtained by implementing the proposed technique on two standard iris databases, IITD and MMU, lucidly outline the promising performance of the astroid shaped feature extraction resulting in a significant increase in rate of recognition accompanied by considerably lower number of features for iris recognition.