Distributed fingerprint enhancement on a multicore cluster

Abstract

Fingerprint enhancement is a crucial step in fingerprint recognition. The accuracy of the recognition algorithm directly depends on the accurate extraction of features which is achieved through a series of image enhancement steps. Unfortunately, the fingerprint enhancement process consists of a series of computationally expensive image processing techniques. This results in slow recognition algorithms. Researchers have examined ways of improving the performance of fingerprint enhancement algorithms through parallel processing. The majority of such techniques are architecture- or machine-specific and do not port well other platforms. We propose a cheaper and portable alternative through the utilization of mixed-mode distributed and parallel algorithms that make use of multicore clusters for processing strength. We tackle a few design concerns encountered when distributing image processing operations. One such concern is dealing with pixels along the borders of the partitioning axis. The other is distributing data that needs to be processed in blocks rather than pixel-wise.