Iris Authentication Utilizing Co-occurrence Matrices and Textile Features

Biometric features have attracted great attention for numerous security applications. Border control, forensics, access control, banking are a few examples where biometrics authentication is applicable. Iris constitutes one of the best suited human physical traits for such applications due to its uniqueness and permanence over time. In this paper, a new iris authentication method is proposed based on gray level co-occurrence matrices of the normalized iris images. The co-occurrence matrices are calculated for four different directions and are utilized for the extraction of textural Haralick features such as energy, contrast, variance, entropy, homogeneity and correlation. This procedure is applied both to the whole image and to segments of it to evaluate the contribution of more localized information to the methodology. For the evaluation of the iris, the aforementioned features are utilized by a Random Forest classifier that judges the authenticity of the user. In order to further enhance the presented method’s performance, both left and right eye irises are evaluated by the classifier and their authentication probabilities are fused forming a robust algorithm.