Classification of IR expressive face images from extracted vascular network

Vascular networks in infrared faces are created due to the blood flow under the skin. Variations in blood flow in the blood vessels cause temperature difference, which produces the vascular networks. This paper deals with binary classification of various infrared facial expressions using vascular network. The classification has been performed using Support Vector Machine classifier on five types of expression where facial features are extracted with the help of Uniform LBP and PCA. Experiments have been conducted on two infrared face datasets: one is our own captured dataset and another is USTC_NVIE database. Experiment results reveal that uniform LBP generate more accuracy than PCA and maximum accuracy is obtained using our own dataset.