Analyses on Multi-sensor Fingerprint Enhancement and Indexing

Wei Zhou, Yong Ding

Abstract: Fingerprint biometric has been widely applied in both forensic law enforcement and security applications, and the most common application is access control. In recent years, we have witnessed the development of touchless fingerprint acquisition technology, which can generate 3D representation of fingerprints. Although some research has been carried out on 3D fingerprint feature extraction (e.g. 3D minutiae) and identification, it is computationally expensive and time-consuming to compare two sets of 3D minutiae in real time. Therefore, the common way to develop an Automatic Fingerprint Identification System (AFIS) using 3D fingerprints is to unwrap 3D fingerprints into 2D equivalent images, then the existing algorithms for 2D fingerprint processing can be utilized. In this paper, we present some observations of 2D to 3D (unraveled 2D equivalent images from 3D model) fingerprints recognition based on a publicly available database. We tested the performance of 2D to 3D fingerprint verification using enhanced unraveled 2D equivalent images. The results show that the cropped enhanced 3D images based on singular points can achieve the best performance. Since enhanced fingerprint images can result in efficient fingerprint indexing, we also made some analyses on the hash bit selection of MCC indexing scheme using the cropped enhanced 3D images based on singular points.