Multispectral Palmprint Recognition based on Fusion of Local Features

One of the most serious challenges faced by the present technological world is creating a personal secured identity. Biometrics is presented as an effective solution to resolve these challenges. But this progress is a double-edged knife that allows malicious persons to reproduce biometric modalities. To overcome this problem, we propose a novel approach for palmprint fusing multiple features. These features are provided from multi-spectral images with 940nm, which allows the extraction of information under the skin of the palm. This information is impossible to reproduce. It is noted that in these images, the gray scale information is capital. In this context, Compound Local Binary Pattern (CLBP) can be an appropriate approach to construct this system, because the CLBP add an extra bit for each P bits encoded by LBP corresponding to a neighbor of the local neighborhood, in order to construct a robustious feature descriptor that exploits both the sign and the inclination information of the differences between the center and the neighbor gray values. The effectiveness of proposed approach has been verified on Casia Multi-Spectral database. The experimental results show that the recognition rates are reliable and the optimal recognition rates can reach 100% for left and right palms.