Article
Evaluation of LBP-based descriptors for iris recognition based on learning vector quantization classifier under a multi-core platform
Autor
Campos, Hernán
Hernández-García, Ruber
Barrientos, Ricardo
Institución
Resumen
In recent years, iris recognition techniques have focused the attention on the biometric area. Particularly, the performance of an iris recognition system depends on the representation of the texture of the images. In this sense, Local Binary Patterns (LBP) have emerged as one of the most prominent and widely studied local texture descriptors. This paper aims to evaluate the use of different LBP-based descriptors for iris recognition by using a Learning Vector Quantization Classifier (LVQ). Our study shows the first analysis of the performance of LVQ classifier on different LBP-based descriptors for iris recognition. Besides, the proposed method evaluates the impact of a parallel implementation of the system under a multi-core platform. The evaluation carried out on the CASIA-Iris-Interval database, shows that LVQ classifier is an effective alternative for iris recognition using LBP-based descriptors. The proposed method achieved a recognition rate of 98.36% and obtaining a speed-up of 24.81x.