Designing palmprint based recognition system using local structure tensor and force field transformation for human identification

This paper presents an efficient palmprint based automatic recognition system. It uses local structure tensor to extract features from the palmprint. Each enhanced palmprint image has been divided into sub-images and features are obtained based on local properties within the sub-image. Force field transformation is used to emphasize the texture of the palmprint and the chosen dominant orientation pixels are used for feature extraction to reduce the effect of noise. Structure tensor values of the dominant orientation pixels within a sub-image are averaged to form the tensor matrix for the sub-image. Eigen decomposition of each tensor matrix is used to generate the feature matrix. Euclidean distance between feature matrices of two palmprints has been used to make the decision on matching. The system has been tested on three databases viz. IITK, CASIA and PolyU and achieved high accuracy. It has also been compared with two best known systems using PolyU database and it has performed better than these two systems.