Title:	Pore detection in fingerprints based on image subtraction and anisotropic diffusion filtering
Authors:	Rodrigues, Emily S. Borges, Vinícius Ruela Pereira
metadata.dc.contributor.affiliation:	University of Brasilia, Department of Computer Science University of Brasilia, Department of Computer Science
Assunto::	Impressões digitais Poros Difusão anisotrópica
Issue Date:	2019
Publisher:	IEEE
Citation:	RODRIGUES, Emily S.; BORGES, Vinícius R. P. Pore detection in fingerprints based on image subtraction and anisotropic diffusion filtering. In: 2018 IEEE INTERNATIONAL CONFERENCE ON SYSTEMS, MAN, AND CYBERNETICS (SMC), 2018, Miyazaki, Japan. Proceedings [...]. [S. l.]: IEEE, 2019. DOI: 10.1109/SMC.2018.00355.
Abstract:	We describe a methodology for extracting and identifying pores in high resolution fingerprint digital images. The key strategy is based on an image subtraction between two fingerprint images with different smoothness levels. The anisotropic diffusion filter is employed to obtain such smoothed fingerprints, in which the pores are preserved in the first fingerprint, while the second one only presents ridges and valleys, but pores are blurred. The subtraction procedure results in a difference image, in which the pores are characterized by the lower magnitudes. After that, we perform a histogram equalization for enhancing pores and a global thresholding to obtain the pores as binary regions. Finally, such binary image is post-processed for removing false pore detections. Experiments were conducted using the PolyU HRF fingerprint image set and the results showed that the proposed methodology outperformed other filtered-based pore extraction methods considering the true and false pore detection rates.
metadata.dc.description.unidade:	Instituto de Ciências Exatas (IE) Departamento de Ciências da Computação (IE CIC)
DOI:	10.1109/SMC.2018.00355
metadata.dc.relation.publisherversion:	https://ieeexplore.ieee.org/document/8616351
Appears in Collections:	Trabalhos apresentados em evento

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.