21-03-2012, 02:16 PM
On the possibility of fingerprint identification by pores detection in 500 dpi images
On the possibility of fingerprint identification by pores detection in 500 dpi images.pdf (Size: 96.08 KB / Downloads: 40)
Introduction
The fingerprint is more used biometric process.
Fingerprint never change (permanence propriety) and
don’t exist two fingerprints same (uniqueness propriety).
In a fingerprint (figure 1), white curves are called valleys
and dark curves are called ridges [1-3]. Galton [3] defines
the first system for fingerprint classification (level 1) and
the minutiae based identification (level 2) as well. The
level 1 considers five different classes (right loop, left
loop, whorl, arch and unknown). Level 2 are useful for
personal identification. On good acquired conditions,
level 2 is very efficient. However when acquisition is
poor or presents no conditions for establish identification,
it is necessary use pores (level 3).
Proposed Method
An approach for detect pores in 1000 dpi images is
proposed by Jain et al. [1]. We propose here a new one
for 500 dpi fingerprint images. In such resolution it must
be much more sensible for pore detection. The used
approach considers seven steps: (1) Thresholding and
background removal; (2) Determination of the local
orientation; (3) Adjustment of the Directional Field; (4)
Local Gabor filter; (5) Combination of original and
filtered images; (6) Pores confirmation; and (7) Counting
and storage.
The first step removes noises and part of the used images
that do not presents the patterns of ridges and valleys. In
this process each portion of 16x16 pixels of image on
analysis, with 8 pixels overlap, is automatically submit to
a conditional thresholding [6].
Conclusion
This work presents a new method to identify pores in 500
dpi fingerprint image. This is based on Gabor filter, and
linear combination. Table 1 shows the possibility of the
use of level 3 features, to individual identification using
500 dpi fingerprint image because for person
identification only 20 to 40 pores are necessary [1] and in
a single fingerprint image more then 1400 pores can be
found. That is, the here presents approach turns possible
the use of pores for identify a person in automated
biometric system using the previous acquired Brazilian
databases.