A Performance Evaluation of Hashing Techniques for 2D and 3D Palmprint Retrieval and Recognition

Biometrics is one of the most effective solutions for personal authentication. With the wide use of biometrics technology, the scale of the biometrics database is increasing continuously. In this situation, fast matching and retrieval techniques become more and more necessary for large-scale biometrics recognition. Recently, palmprint recognition has become one of emerging biometrics technologies. Both low-resolution palmprint recognition and 3D palmprint recognition technologies were proposed and can be used for many civilian applications. However, the research on fast low-resolution palmprint and 3D palmprint retrieval is very preliminary. Hashing is one of the most popular techniques for image search due to its fast-computational speed and low storage cost. In this paper, we conduct a performance evaluation of hashing techniques for 2D and 3D palmprint retrieval and recognition. Several representative hashing methods are selected for performance evaluation including four unsupervised hashing methods, four supervised hashing methods, and four deep hashing methods. Experimental results demonstrate that some hashing methods such as column sampling based discrete supervised hashing (COSDISH) can obtain promising results.