Fusing Discrete Cosine Transform and Multi-level Center-Symmetric Local Binary Pattern Features for Periocular Recognition

Abstract

A novel method by fusing the features of discrete cosine transform (DCT) and multi-level center-symmetric local binary pattern (CS-LBP) is proposed for periocular recognition in this paper. Because that CS-LBP is used to extract features from the images only once, by which the extracted texture features are not adequate to represent the periocular images, we employ multi-level CS-LBP to extract more abundant and informative texture features for more times to get the spatial features. The primary information of the periocular image was centralized in a small number of DCT coefficients which were used as the frequency features of the image. The periocular image was divided regularly into small regions from which histograms were computed and concatenated into a spatial global histogram used as descriptor vector of the periocular image. Then the DCT features and the CS-LBP features were fused posterior to the normalization. Experimental results on ORL face database, AR face database and Morph periocular database demonstrate that the proposed method outperforms DCT or LBP features for periocular recognition.