Defocused iris image restoration based on spectral curve fitting

Despite the rapid development of iris image capturing system and recognition algorithms, defocused iris images occurs a lot due to the restriction of optical system's depth of field. To take advantage of the useful texture information for iris recognition in more unconstrained environment, we proposed a scheme for iris image deblurring based on fitting ellipse curve in the frequency spectral image. To get the parameters for Point Spread Function(PSF) initialization, we first calculate the frequency spectral image of defocused iris images, and then fit the ellipse with Hough Transform to get the defocus estimation. Blind deconvolution is chosen as the restoration method in which the PSF is refined through iteration. Experiments with both artificial data and real data are conducted and the results demonstrated the effectiveness of our method to restore defocused iris images.