Palmprint Classification via Filter Faces and Feature Extraction

Palmprint classification is a popular method for today’s biometrics, and it can be combined with iris or fingerprint to identify a person’s identification. In this paper, we propose a novel method for palmprint classification by using filter faces and feature extraction. We extract such features as histogram of oriented features (HOG) and local binary patterns (LBP). We select to use both nearest neighbor (NN) classifier and collaborative representation classifier (CRC) for palmprint classification. Experiments demonstrate that filter faces always improve the classification accuracy for palmprint classification no matter what features are used and how much noise is present in palmprint images. Based on our experiments, the HOG features are very robust to noise whereas the LBP features are very sensitive to noise. As a result, it is preferable to use the HOG features for noisy palmprint images and the LBP features for clean palmprint images to improve classification accuracy.