Iris Recognition Using Discrete Cosine Transform and Relational Measures

Iris is one of the most currently used biometric trait as it has random discriminating texture which does not change over a person’s lifetime. It is unique for all individuals, even for twins and the left and right eyes of the same individual. This paper describes an iris recognition system which includes phases like segmentation, normalization, segregating unwanted parts like occlusion, specular reflection and noise, enhancement, feature extraction and matching. A new feature extraction technique based on Discrete Cosine Transform (DCT) for iris recognition has been proposed. Also, a new fusion strategy of fusing the dissimilarity scores of the two feature extraction techniques (DCT and Existing Relational Measures) has been proposed to get better performance. The experimental results have been shown on large databases like publicly available CASIA-4.0 Interval and CASIA-4.0 Lamp and self-collected IITK, and the proposed fused approaches have achieved encouraging performance results.