Efficient iris recognition through curvelet transform and polynomial fitting

摘要

This paper presents a new feature descriptor for iris recognition. The descriptor makes use of the wedge-shaped sub-bands of the curvelet transform, which allow it to cover the complete frequency spectrum, to characterize the iris images. The texture present in each of the curvelet sub band is further represented by fitting a 2D polynomial of appropriate degree. Coefficients of polynomials fitted to each of the curvelet sub-bands are further collected to form the complete feature vector. The proposed approach is investigated with benchmark IITD and CASIA-v4-Interval iris databases to prove its usefulness. Results calculated in terms of area under receiver operator characteristics (ROC) curves (AUC) and equal error rates (EER) clearly display the outperforming nature of the proposed descriptor.