FilterbankBased Fingerprint Matching

1
Filterbank-Based Fingerprint Matching

• Kitiwat Limmongkol
• Cheng-Yu Yang
• Columbia University
• Spring 2006

E6886 Topics in Signal Processing Multimedia
Security System
Final Project Proposal Presentation
May 10, 2006
2
Presentation Outline
E6886 Topics in Signal Processing Multimedia
Security System

• Algorithm part
• System diagram
• Demo
• Result part
• Table of distance
• Percent of accuracy
• Note
• Summary part
• Discussion
• Conclusion
• Reference

I
II
III
3
Four Main Steps in feature extraction algorithm
E6886 Topics in Signal Processing Multimedia
Security System

• 1.Determine a reference point and region of
  interest for the fingerprint image.
• 2.Tessellate the region of interest around the
  reference point.
• 3.Filter the region of in eight different
  directions using a bank of Gabor filters.
• 4.Compute the average absolute deviation from the
  mean (AAD) of gray values in individual sectors
  in filtered images to define the feature vector
  or the FingerCode.

I
II
III
4
System Diagram
E6886 Topics in Signal Processing Multimedia
Security System
I
II
Fig1 System diagram of the fingerprint
authentication system
III
5
Different points in a fingerprint
Finding a reference point1. choose the reference
point manually.2. compute the appropriate
orientation field and use identification
masks.3. Poincare Index method.4. Method use in
the paper 1.
I
II
III
6
Poincare Index method

• 1. try to find directional field
• 2. Detect the singular point
• (1) Estimate and smooth the directional field
  of input fingerprint image.
• (2) In each block(88),we compute the Poincare
  index. The Poincare index is compute as follows

I
II
III
7
Poincare Index method (Cont.)

• (3) If the Poincare Index is ½,then this block
  is the core block. The center of this block is
  the core point. If more than two core points are
  detected ,go to step 1.

I
II
III
8
Sample Fingerprint database

• Fingerprint samples were scanned with Compaq
  DFR-200 scanner at 500 dpi, http//www.neurotechno
  logija.com
• We used 14 pairs, 28 fingerprints
• (2 fingerprints per person).

I
II
III
9
Fingercode
E6886 Topics in Signal Processing Multimedia
Security System
I
II
III
10
Demo

• Please wait for a second.

I
II
III
11
Presentation Outline
E6886 Topics in Signal Processing Multimedia
Security System

• Algorithm part
• System diagram
• Demo
• Result part
• Table of distance
• Percent of accuracy
• Note
• Summary part
• Discussion
• Conclusion
• Reference

I
II
III
12
Table of Distance
E6886 Topics in Signal Processing Multimedia
Security System
Link to excel file
I
II
III
13
Percent of Accuracy
Person Identification2/14 people fail to
distinguish from the others. 85.71 success
Database Accuracy3/28 fingerprints cannot match
with their pair 89.29 success
I
II
III
14
Note
E6886 Topics in Signal Processing Multimedia
Security System

• We perform matching 2828 764 times.
• H2, M1, B1 fingerprints created false acceptance
  of C2, F1, F2 respectively.

I
II
III
15
E6886 Topics in Signal Processing Multimedia
Security System
Presentation Outline

• Algorithm part
• System diagram
• Demo
• Result part
• Table of distance
• Percent of accuracy
• Note
• Summary part
• Discussion
• Improvement
• Conclusion

I
II
III
16
Discussion Example of reject images

• H2

C2
I
B1
F2
II
III
17
Improvement
E6886 Topics in Signal Processing Multimedia
Security System

• Also use frequency domain to find the reference
  point due to high frequency at the center point.
• Use Gaussian low pass filter to remove noise from
  the test images.
• Crop image after find the reference point in
  order to avoid zero problem from rotation and
  improve processing time
• Simple GUI for matching fingerprints.

I
II
III
18
Conclusion

• Our fingerprint database size is small compare
  with MSU_DBI database (14 people compare with 167
  people).
• Therefore, the result is not consistency and can
  be changed when we run our system with the large
  and standard database. (Hopefully, the percentage
  of accuracy will increase).
• We have to figure out how to compare the
  fingercode when an appropriate region of interest
  could not be constructed and quality of image was
  poor (center of fingerprint locates near the
  edge, etc.).

I
II
III
19
References

• 1 A. K. Jain, S. Prabhakar, L. Hong, and S.
  Pankanti, Filterbank-Based Fingerprint
  Matching, IEEE Trans. Image Processing, vol. 9,
  no. 5, pp. 846-859, 2000.
• 2 A. K. Jain, L. Hong, S. Pankanti, and R.
  Bolle, An identity authentication system using
  fingerprints, Proc. IEEE, vol. 85, pp.
  13651388, Sept. 1997.
• 3 R. O. Duda and P. E. Hart, Pattern
  Classification and Scene Analysis, New York
  Wiley, 1973
• 4 Adhiwiyogo, S. Chong, J. Huang, W. Teo, Final
  Report 18-551 (Spring 1999) Fingerprint
  Recognition Group Number 19Markus,
  http//www.ece.cmu.edu/ee551/Old_projects/
  projects/s99_19/finalreport.html

I
II
III