Abstract:In order to improve the correct recognition rate for poor quality iris images

a sectored-based method for iris recognition was proposed. Not like the conventional recognition method

the algorithm divided the iris image into different sectors

recognized different sectors separately and combined their results using Bayesian fusion based on sparsity concentration index value. The low-rank matrix of each sector was selected as iris feature

and sparse representation based classification was used as classifier. Because different regions of the iris had different qualities

the algorithm can reduced the impact of poor quality regions on the recognition result. Experiment results on CASIA-Iris-Interval and IIT Delhi V1 iris database showed that the proposed algorithm achieved better correct recognition rate and had strong robustness for poor quality iris images. The studies of the number of sectors and training images demonstrated that

eight of the sector number was suggested and a test iris image could be well represented by seven training image from the same class.