Authors: Diwakar Agarwal; Atul Bansal
Addresses: Department of Electronics and Communication, GLA University, Mathura, Uttar Pradesh, India ' Department of Electronics and Communication, GLA University, Mathura, Uttar Pradesh, India
Abstract: Extraction of fingerprint sweat pores is a critical step in those applications which are based on highly secured features. Pores are varying in scale (size) and evenly distributed along the ridges. It is the main challenge to design a technique which determines the pores of different sizes in the fingerprint image. In this paper, pore extraction algorithm is proposed for high-resolution fingerprint images which utilised multiscale γ-normalised Laplacian of Gaussian (LoG) filter. A block-wise approach is implemented in which each region is filtered at multiple scale values. Scale space theory is applied and candidate pixels of high negative response are identified through local maxima approach. The efficacy of the proposed algorithm is tested by measuring average true detection rate (TDR) and average false detection rate (FDR). Results of the proposed algorithm achieve average TDR and average FDR values as 82.89% and 21.2% respectively which are better in comparison to the state-of-art techniques.
Keywords: automatic scale selection; biometrics; fingerprint; local maxima; pores.
International Journal of Biometrics, 2020 Vol.12 No.3, pp.317 - 336
Accepted: 15 Dec 2019
Published online: 24 Jun 2020 *