Research On Fast Iris Localization And Iris Image Quality Assessment Algorithms

Iris localization is very important for iris recognition system. In previous research, the accuracy and the speed of locating algorithm are constrained by each other. Most accurate iris locating algorithm can't satisfy the time requirement of iris recognition system. In this paper, we propose a nonlinear method using the nonlinear data which are the candidate edge pixels of iris to fit the iris boundaries by least squares combined with vertical projection of the gray scale level. The experiments of the proposed algorithm has been compared among other implemented algorithms on the CASIA iris image database, and the experiment results demonstrated that the proposed algorithm has the fastest speed than other algorithms while ensure the locating accuracy.The quality of iris image also is very important for iris recognition system. In practice, the error matching often causes by the terrible quality of iris images. In order to avoid this problem we must evaluate the quality of iris image and select only the quality fitting our request and eliminate the bad quality one. However, not anymore efforts are done for this problem and the current algorithms of image quality assessment are still limited. In this paper, I propose a kind of algorithm based on Laplacian sharpening combined with FFT (Fast Fourier Transform). This algorithm not only can evaluate the quality of iris image well but also classify bad quality into 3 different situations including defocused image, motion blurred image, eyelid and lash occlusion, in order to prompt the users or the iris system to correct the errors.