Iris recognition system based on fuzzy local binary pattern histogram and multiple classifiers

The local binary pattern (LBP) histogram is known as a potent and attractive texture descriptor that shows brilliant results. A number of new extensions to LBP-based texture descriptors have been proposed that focus on noise intensity enhancement. In iris recognition system, one of the main challenges is to improve the robustness of image brightness and noise changes by using different coding or thresholding schemes. This paper proposes the fuzzy local binary pattern (FLBP) approach to extract iris features at the early recognition stage. FLBP has several advantages including ease of execution and robustness under changing image conditions. The proposed system provided a significant improvement in recognition performance as it has the benefit of the strength of the histogram that provides the great majority of local context information. The proposed iris detection design uses support vector machine (SVM), linear discriminant analysis (LDA), and k-nearest neighbor (KNN) classifier for classification. Experiments with two challenging iris databases (CASIA-VI and CASIA-V4) were performed to determine the usefulness of the approaches. The method has been evaluated against existing techniques and the proposed approaches have produced excellent results.