An approach to aligning regions of interest on infrared palm-dorsum images captured in contactless scenarios

This paper presents an approach for solving a highly challenging problem in the preprocessing of hand-based biometrics, namely restoring IR palm-dorsum images captured in contactless scenarios and aligning regions of interest (ROIs) in the same regions on the various restored infrared (IR) palm-dorsum images captured from the same palm-dorsum. IR palm-dorsum images captured in contactless scenarios substantially increase the user-friendliness, security and sanitation of hand-based biometrics. However, images captured in contactless scenarios typically exhibit rotation, translation, scale and shear geometric transformations. These geometric transformations significantly reduce the accuracy of hand-based biometrics. An approach used to restore and align ROI on an IR palm-dorsum image is proposed to solve this problem. The proposed approach is based on a two-dimensional affine image transformation scheme used to restore the geometric transformations of the images, and an ROI alignment method based on finger-web positions. Thus, the proposed approach improves the user-friendliness, security, sanitation and accuracy of hand-based biometrics based on features extracted from the ROI. The principal characteristics of the proposed approach are no prior information on IR images being necessary, and no parameters being required to be preset. The experimental results indicated the effectiveness and feasibility of the proposed approach. The proposed approach improves accuracy, user-friendliness, security and sanitation and extends the application of hand-based biometrics for use in security access control systems.