Effective Iris Recognition for Distant Images Using Log-Gabor Wavelet Based Contourlet Transform Features

摘要

Distant iris recognition has become an active research topic in bio-metric as well as computer vision, but it is still a very challenging problem. In order to solve it effectively, we propose a novel framework by utilizing Log-Gabor wavelet based Contourlet transform (LGCT) feature descriptor with an effective kernel based extreme learning machine (KELM) classifier. The experiments are conducted on CASIA-v4 which is a typical database of distant iris images. It is demonstrated by the experimental results that our proposed LGCT features are quite effective for distant iris recognition and the highest accuracy can arrive at 95.93% when they are fused with the convolutional neural networks (CNN) and gradient local auto-correlations (GLAC) features together.