Entropy based occlusion removal with DTCWT and DCT as feature extractor for iris recognition

The iris pattern of human eye is considered to be unique and distinct that it plays an important role as reliable biometric system for authentication purposes. Iris recognition has become a challenging issue under varying lighting or contrast conditions and due to occlusions such as eyelashes. Therefore, in this paper we have proposed a novel technique for eliminating the eyelashes, namely Entropy based Occlusion Removal which is applied after iris localization. Also a feature extractor based on Dual Tree Complex Wavelet Transform (DTCWT) is used for extracting shift-invariant features in combination with Discrete Cosine Transform (DCT). Binary Particle Swarm Optimization (BPSO) is used for feature selection that selects the optimum features extracted by DTCWT+DCT. The Euclidean distance classifier estimates the similarity between the testing and training images. The experiments conducted on IITD and CASIA iris databases depict promising performance of the proposed technique.