Shifted 1D-LBP based ear recognition system
Mohamed Benouis, Meryem Regouid · 2020
Human ear recognition presents a rich and stable structure which may be considered as a perfect source of information for identification or verification purpose. Ear is largely used in recent years for recognition purpose. In this study, the authors designed and implemented a novel ear based biometric identification system based on shifted ID-LBP. First, after ear image pre-processing and being converted to ID space, ID-LBP is used to reduce the dimensions of each ear image vector. In ID-LBP, a histogram bin is constructed to encode the temporal and spatial information in order to obtain more discriminative features. We perform our method on USTB public datasets, the accuracy achieves 97.04% and 100% with the nearest neighbor classifier. A comparison with the state-of-the-art methods show that the proposed method performs better.