finger print pore extraction methods

1

ARVIND S. SARDAR 10.10.2012 Adaptive fingerprint pore modelling and extraction

ADAPTIVE FINGERPRINT

PORE MODELING

AND

EXTACTION

2


Introduction

• Biometrics:

Biometrics is a study of methods for uniquely recognizing

humans based upon one or more intrinsic physical or

behavioral characteristics.

• Biometrics can be sorted into two classes:

• Physiological

Examples: face, fingerprint, hand geometry and iris recognition

• Behavioral

Examples: signature and voice

3


Introduction• Properties of biometrics 1. Universality Every person should have the biometric characteristic2. Uniqueness No two persons should be the same in terms of the biometric characteristic3. Permanence The biometric characteristic should be invariant over time4. Collectability The biometric characteristic should be measurable with some (practical) sensing device5. Acceptability One would want to minimize the objections of the users to the measuring/collection of the biometric6. Circumvention which reflects how easy it is to fool the system by fraudulent methods.

4


General Biometric System

Database

BiometricSensor Feature Extraction

BiometricSensor Feature Extraction

Matching

ID : 8809

Authentication

Enrollment

ResultAuthentication Result

http://images.google.com/imgres?imgurl=www.informatik.uni-stuttgart.de/ipvr/vs/personen/mecki/fingerprint.gif&imgrefurl=http://airu.persianblog.com/&h=199&w=193&prev=/images?q=fingerprint&svnum=10&hl=en&lr=&ie=UTF-8&oe=UTF-8&sa=N



5


FINGERPRINT AS A BIOMETRIC

• A fingerprint is an impression of the friction ridges, from the surface

of a fingertip.

• It is used for personal identification

• Easy in acquisition

• High matching accuracy rate

• Do not change over time

• Dominate biometric market by accounting for 52% of authentication

systems

6



Fingerprint representation The types of information that can be collected from a fingerprint’s

friction ridge impression can be categorized as level 1, Level 2, Level 3.

• Level 1

The fingerprint pattern exhibits one or more regions where the ridges lines assume distinctive shapes characterized by high curvature, frequent termination.

• Level 2

ridge ending and ridge bifurcations

• Level 3

Fine intra ridge details

7




8



“Most of fingerprint identification systems (like AFIS)

rely on minutiae (Level 1&2) only. While this information

is sufficient for matching fingerprints in small databases,

it is not discriminatory enough to provide good results

on large collections of fingerprint images.“

[M. Ray, P. Meenen, R. Adhami - “A Novel Approach to Fingerprint Pore Extraction“, IEEE, Mar. 2005]

9



• both show a bifurcation at the same location

– Examination based on Level 1&2 features – match– In combination with Level 3 features (e.g. relative pore position) – no match

10


Physiology – Fingerprint formation

• Fingerprints begin forming on the fetus 13th week of development

• Ridge units are fusing together as they grow forming ridges

• Each ridge unit contains a pore which originates from a sweat gland from the dermis

• Pores are only found on ridges not in valleys

11


Physiology – Some facts

• typical fingerprint: 150 ridges

• A ridge ~ 5 mm long contains appr. 10 ridge units

• Ridge width: ~ 0.5 mm

• Average number of pores / cm ridge ~ 9-18 pores

• Pores do not disappear, move or generate over time

[Ashbaugh, D., Quantitative-Qualitative Friction Ridge Analysis, 1999, CRC Press]

[Locard, Les pores et l'identification des criminals, Biologica, vol.2, pp. 257-365, 1912]

12


Pore Extraction methods

1. skeleton-tracking-based methods

- First binaries and skeletonize the fingerprint image and then track

the fingerprint skeletons.

- Computationally expensive.

- very sensitive to noise.

- work well on very high resolution fingerprint images.

2. Filtering-based methods

- filter fingerprint images

13


Isotropic pore models

Invariant with respective to direction

1. Ray’s Model:-

which is used 2-dimensional Gaussian functions for pore extraction.

2. Jain’s model:-

Jain proposed to use the Mexican hat wavelet transform to extract pores based on the observation that pore regions.

3. DoG Model:- (Difference of Gaussian filter )

Is to use a band-pass filter to detect circle-like features.

14


Proposed system

Propose to develop a novel dynamic anisotropic pore model

which describes the pores more flexibly and accurately by

using orientation and scale parameters and an adaptive

pore extraction method can detect pores more accurately

and robustly.

15


Dynamic anisotropic pore model (DAPM)

• Previous models are isotropic and static (uses unitary scale)

• This new pore model has two parameters to adjust scale and orientation,

• These two parameters are adaptively determined according to the

local ridge features (i.e. ridge orientation and frequency)

• DAPM is defined

Eq. (1) is the Reference Model (i.e. the zero-degree model)

Eq. (2) is the rotated modelHere, is the scale parameter which is used to control the pore size. It can be determined by the local ridge frequency. Is the orientation parameter which is used to control the direction of the pore model.

16


Adaptive pore extraction method

• Pore extraction is essentially a problem of object detection.

• DAPM parameter estimation:-

To instantiate the pore model initialize two parameters orientation and scale.

- Orientation parameter :-Set the local fingerprint ridge orientation

- Scale parameters :- Use the maximum valid pore scale

• Implementation issue:-With estimated parameter an adaptive pore model can be instantiated for each pixel and apply to matched

Filter to extracted pore from the fingerprint image.

- computational cost:-

Calculations as pixel wise way

17



• Implementation issue:-

- Accurate estimate

Difficult to get accurate estimate by pixel wise

• To deal with these issue, propose Block wise approach

• Three kinds of blocks on fingerprint image

1) Well-defined blocks

2) Ill-posed blocks Foreground fingerprint

region

3) Background blocks

18



• well- defined block: It is able to directly estimate a dominant ridge orientation and a ridge frequency.• Ill-posed block: There is not a dominant ridge orientation but the ridge frequency can be estimated by interpolation of the frequencies on its neighboring blocks.

19



• Pore Extraction algorithm:

20



• Partition:

The first step is to partition the fingerprint image into a number of blocks, each being a well-defined block, an ill-posed block or a background block

• Ridge orientation and frequency estimation:

The ridge orientation field of the fingerprint image is calculated. Meanwhile, the mean ridge frequencies on all foreground blocks are estimated, which form the ridge frequency map of the fingerprint image.

• Ridge map extraction

The binary ridge map of the fingerprint image is calculated

• Pore detection:

The foreground fingerprint blocks are processed one by one to detect pores

on them

21



• Post Processing

Record the extracted pores by recording the coordinates of their mass centers

22


Thank you

finger print pore extraction methods

Education

fingerprint pore extraction

biometric fingerprint

fingerprint skeletons

fingerprint pattern

adaptive pore ext

relative pore

pore extraction methods1

facts typical fingerprint