Joint Multiple-type Features Encoding for Palmprint Recognition

Palmprint contains rich unique features such as lines and textures for reliable personal authentication. There have been a number of methods proposed for palmprint recognition in recent years. However, most existing palmprint feature descriptors only extract single-type features, which usually can not completely represent the multiple information of a palmprint. In this paper, we propose a novel joint multiple-type features encoding method, which jointly extracts and encodes the important direction and texture features for palmprint recognition. Specifically, we first extract both the dominant direction and the gradient change data to respectively describe the direction and texture features of a palmprint. Unlike the existing methods that directly extract features from single pixel, we extract both the direction and texture features among the local patch by using a majority voting scheme so that the extracted multiple-type features are more accurate and reliable. Finally, we jointly encode the multiple-type features into decimal feature code, and pool them into block-wise histogram feature descriptor for palmprint representation and recognition. Extensive experimental results on the baseline palmprint databases, including the CASIA, IITD and TJU databases, demonstrate the competitive effectiveness of the proposed method.