Infrared Face Recognition Based on Personalized Features Selection of LBP

The compact and discriminative feature extraction is vital for infrared face recognition. This paper proposes a personalized feature selection algorithm for infrared face recognition. Firstly, LBP operator is applied to infrared face for texture information. Secondly, for each subject, a two-class training problem is constructed by one to other means. Then, based on two-class discriminative ability, we adaptively select a personalized subset of features from LBP for each subject. Finally, the nearest neighbor classifier based on chi-square distance is utilized to get final recognition result. The experimental results show the personalized feature selection is effective in useful information extraction for infrared face recognition, which outperform the state of the art methods based on LBP.