Iris Recognition Under Partial Occlusion Based on Non-negative Sparse Representation Classification

摘要

To improve the reliability and accuracy of personal identification based on iris under partial occlusion, this paper proposed a non-negative dictionary sparse representation and classification scheme for iris recognition. The non-negative dictionary includes the Log-Gabor feature dictionary extracted from normalized iris image. The use of Log-Gabor makes the occlusion dictionary compressible, and can reduce the computational cost. Experiments on UBIRIS iris database demonstrated the effectiveness of the proposed Log-Gabor based non-negative sparse representation classification.