1 Local Correlation-based Fingerprint Matching Authors: Karthik Nandakumar, Anil K. Jain Source: To...

Preview:

Citation preview

1

Local Correlation-based Fingerprint Matching

Authors: Karthik Nandakumar, Anil K. JainSource: To Appear in Proceedings of

ICVGIP, Kolkata, December 2004Speaker: Shu-Fen Chiou (邱淑芬)

2

Outline Introduction Proposed Method Experimental Results Conclusions Comment

3

指紋特性 : 唯一 / 萬人不同的關鍵

性 特徵本身之永久不變 事後追查的能力 特徵取得之變動性

4

指紋辨識

Template Minutiae Query Minutiae

5

Proposed Method Minutiae Extraction Fingerprint Alignment Local Correlation-based Matching

6

Minutiae Extraction

7

Minutiae Extraction

8

Fingerprint Alignment

QTQT

TQ

qq

QTkqkqq

, tk, tt

Q

T

and in points theofmean the: ,

toin points theansformsnt that trdisplaceme the:

of transposeconjugatecomplex the:

and oflength minimal: ) , ... ,1(

) 1(

query in the ridge ingcorrespondon points ofset a:

ein templat points ridge ofset a:

*

9

Fingerprint Alignment

TTd

QT

and between error squraed of sum the:

and , using predicted :

cd if

:ingcorrespond possible considered ispair minutia a

30 rotation estimated if

image theofheight theof half direction verticalestimated if

:ingcorrespond possible considerednot ispair minutia a

10

Local Correlation-based Matching

11

Experimental Results

Data : 160 users

12

Experimental Results

13

Experimental Results

14

Conclusions The performance of our algorithm is slightly

inferior to that of the 2D dynamic programming based minutiae matcher, mainly due to the inability to handle fingerprint images of very low quality.

However, integrating the proposed algorithm with the 2D dynamic programming based matching yields a better matcher.

15

Comment

R treedb

Location, relation,…Template

image

Query image

R tree

Match

Recommended