A secure and robust indexing algorithm for distorted fingerprints and latent palmprints

abstract

© 2022 Elsevier LtdNowadays, both fingerprint and palmprint databases are massive and whose size has exceeded millions. Therefore, applying a filtering technique, such as indexing, is vital for automatic fingerprint/palmprint identification systems. Geometric distortion present in fingerprints and the creases in palmprints have a significant drop in recognition accuracy. Moreover, protecting an enrollment template is an important and challenging issue today. Template protection is a technique to convert an unprotected enrollment biometric template to a protected biometric template and is used to prevent the access of illegal users and attackers to the enrolled biometric template. Hence, in this paper, we propose a new feature called the middle of the triangle's side (MTS) derived from each minutiae pair of a triangle-based representation to mitigate the negative effect of the geometric distortion and creases. Furthermore, they can be used as a feature transformation to protect templates. For computing MTSs, we first estimate the quality of input images and then extract reliable minutiae from the input images. After that, we apply the Delaunay triangulation of order k to minutiae for obtaining a triangle-based representation. Then, we calculate the median of sides of each triangle for extracting MTSs. To obtain the direction of each MTS, we use the direction difference between each minutiae pair placed in both vertices of the triangle's side. This makes MTSs robust against fingerprint and palmprint rotation. Afterward, we obtain new triangles by connecting the MTSs of each triangle and weight feature vectors based on the quality of minutiae, and generate indices. Finally, we propose a new feature vector for triangles to increase the recognition accuracy. Experimental results on two public fingerprint databases containing distorted fingerprints, and two public palmprint databases containing latent prints, show that MTSs are more robust than minutiae against geometric distortion. Also, MTSs are secure, and its reason is that minutiae direction and location are changed. In addition, the number of MTSs is more than minutiae, making them more appropriate for prints with a low number of minutiae. Our experimental results show that MTSs and the proposed indexing algorithm can obtain good results for distorted fingerprints and latent palmprints.