Infrared Face Recognition Based on Local Derivative Binary Pattern and Pattern Selection

An infrared face recognition method combining two-order local micro-patterns and patterns selection is proposed in this paper. Firstly, according to local derivative binary pattern (LDBP) encoding, the two-order local directions features are extracted. Secondly, based on supervised learn idea, pattern selection (PS) algorithm is proposed to get the LDBP patterns which are most suitable for infrared face recognition. Finally, to make full use of the space locations information, the partitioning and LDBP histogram are applied to get final features. The experimental results demonstrate combination of LDBP and PS improve the performance of the infrared face recognition, our proposed method better recognition rate with fewer features compared with the method based on LBP and PCA.