TR-LBP: A modified Local Binary Pattem-based technique for 3D face recognition

In this paper, a novel technique has been introduced for 3D face recognition based on the modified local binary pattern extracted from a 3D range image. The new LBP technique is applied to shape index of 3D facial surface data. The novelty of this technique illustrates that a modified local binary pattern termed as Triangular Local Binary Pattern (TR-LBP), which gives new texture representation of 3D facial surface for improvement of facial texture classification performance compared to other variants of LBP. In this paper, authors have also described the TR-LBP technique by extending it on 2D intensity image of same subjects. Entropy-based feature extraction is used for feature vector creation. Further, KNN is used for calculating classification accuracy on two popular 3D face databases: Frav3D and Bosphorous. Here the classifications results are compared with other two existing LBP techniques applied to range images and shape index (SI) form of range image respectively.