Comparative study of 1D-local descriptors for ear biometric system

As an important modality for human identification, Ear based biometric has achieved a relatively mature level of development, as it faces higher challenges surrounded by the real-world applications of biometric technology. One such challenge is extracting a unique template that leads to making a reliable identification task. In the most existing ear biometric approaches, the features were calculated from the 2D and 3D images. In the presented work, we have well investigated the performance of 1D-LBP and its variations (i.e., standard 1D-LBP, shifted-1D-LBP, 1D-Multi-Resolution-LBP, Local Centroid Pattern, Local Ternary Pattern, Local neighbor gradient pattern and 1D-Noise-tolerant local binary pattern) on ear recognition. Typically, the 1D-LBP treats the ear image as a 1D vector where the histograms of the produced image are then used as features to describe a human ear. The experimental results show that the LBP’s in 1D is promising in developing a robust handcrafted feature for ear recognition.