FINGERPRINT IMAGE ENHANCEMENT USING STFT ANALYSIS

1
FINGERPRINT IMAGE ENHANCEMENT USING STFT ANALYSIS

• Sharat Chikkerur, A. N. Cartwright and Venu
  Govindaraju
• Center for Unified Biometrics and Sensors
• University at Buffalo
• www.cubs.buffalo.edu

2
Abstract

• Motivation
• Feature extraction is very unreliable in poor
  quality prints
• Matching accuracy can be improved by inclusion of
  an enhancement algorithm
• Contributions
• New fingerprint image enhancement using STFT
  Analysis.
• Local region of the fingerprint is modeled as a
  surface wave
• Fourier Transforms is treated as a probabilistic
  distribution of surface waves
• The algorithm simultaneously extracts all
  intrinsic images of the fingerprint (orientation
  map, ridge frequency map and foreground map)
• Enhancement also performed in the Fourier domain
• Evaluation
• Objective evaluation performed over FVC 2002 DB3
  database
• Algorithm compares favorably with Gabor Filter
  based approach

3
Outline

• Need for Enhancement
• Literature Survey
• STFT Analysis
• Enhancement Results
• Objective Evaluation

4
Need for Enhancement
High contrast print
Typical dry print
Faint print
Low contrast print
Typical Wet Print
Creases
5
Prior Related Work

• Challenges
• Fingerprint image is non stationary (has dominant
  local orientation and frequency)
• General purpose image processing algorithms are
  not useful
• Traditional operators and filters assume gaussian
  noise model
• Noise in fingerprint images consists mostly of
  ridge breaks
• Contextual Filters
• Existing techniques are based on contextual
  filtering
• Filter parameters are adapted to each local
  neighborhood
• Filters themselves may be spatial or Fourier
  domain based
• Filter parameters in unrecoverable regions can
  be interpolated based on its neighbors

6
Spatial Filtering

• (Yang et.al 1996, Greenberg et. Al 1999) proposed
  local anisotropic filtering
• Filter kernel adapts at each pixel location

• Parameters
• radial extent of the filter
• vector parallel to the ridge direction ridge
  direction
• vector perpendicular to the ridge direction
• , , shape parameters
• In our case, S -2, V 10 , 4,
  2

7
Spatial Filtering (cont.)

• Hong et al, 96/98 proposed the use of Gabor
  filters for enhancement
• Gabor filter has the best joint space-frequency
  localization
• The filter is aligned with the direction of the
  ridges
• Does not handle high curvature regions well due
  to block wise approach.
• Angular and radial bandwidths are constant.

Even Symmetric Kernel
Fourier spectrum showing the localization
8
Fourier Domain Filtering

• Sherlock et al 94, proposed the use of Fourier
  domain filtering
• The image is convolved with a filter bank of
  directionally selective filters
• Image enhanced by selecting a linear combination
  of filter responses
• Has high space complexity, requires estimation of
  core/delta locations
• Watson et al. 94, proposed the use or root
  filtering for enhancement.(Pseudo matched
  filter)
• Does not require the computation of orientation
  images

Root Filtering
Fourier Domain Filtering
9
Traditional Approaches
Local Orientation ?(x,y) Gradient Method
Enhancement Frequency/Spatial
Local Ridge Spacing F(x,y) Projection Based Method
Ratha et al 95
10
Proposed Approach Overview
Region Mask
STFT Analysis
Frequency Image
Fourier domain Enhancement
Orientation Image
Coherence Image
11
STFT Analysis

• Fingerprint image is non stationary, so we
  require both space and frequency resolution time
  frequency analysis
• STFT in 1D
• STFT in 2D

12
Surface Wave Model
Fingerprint ridges can be modeled as an oriented
wave
Surface wave
Local Neighborhoods
Validity of the model
13
Parameter Estimation

• Paradigm The Fourier domain response can be
  viewed as a distribution of surface waves. Each
  term F(r, ?) corresponds to a surface wave of
  frequency 1/r and orientation ?
• We seek to find the most likely surface wave and
  hence estimate the dominant direction and
  frequency
• We can represent the Fourier spectrum in polar
  form as F(r,?) The power spectrum is reduced to a
  joint probability density function using
• The angular and frequency densities are given by
  marginal density functions

14
Ridge Orientation Image
15
Frequency Image
Jain et al 00
16
Region Mask

• The surface wave approximation does not hold in
  the background region
• The region mask is obtained by simple
  thresholding of the block energy image

17
Coherence Image

• Block processing is unreliable in regions of
  high curvature
• Sherlock and Monro 94, relax filter parameters
  near the singular locations
• Estimation of singular point is difficult in
  poor images!
• We use an angular coherence measure proposed by
  Rao and Jain 90

18
Enhancement
19
Additional Enhancement Results
20
Qualitative Comparison
Watson et al, 94
21
Qualitative Comparison(cont.)
Proposed approach
Hong et al, 97 (Gabor filtering)
22
Objective Evaluation

• We evaluated the effect of enhancement on 800
  images from FVC2002 DB3
• The evaluation consists of 2800 genuine test and
  4950 impostor tests
• It can be seen that the matcher performance
  improves with enhancement

23
Software

• Matlab code for our proposed approach is
  available from http//www.eng.buffalo.edu/ssc5
• Matlab code for our implementation of Hong et.
  Al, Watson et. Al is available from
  http//www.cubs.buffalo.edu/

24
Conclusion

• We presented a new enhancement algorithm based on
  STFT analysis
• Advantages of the algorithm
• All the instrinsic images(ridge orientation,ridge
  frequency, region mask) are estimated
  simultaneously from STFT analysis.
• The estimation is probabilistic and is therefore
  more robust.
• The enhancement utilizes the full contextual
  information(orientation,frequency,angular
  coherence) for enhancement.
• The algorithm has reduced space requirements
  compared to more popular Fourier domain based
  filtering techniques.
• We perform an objective evaluation of the
  enhancement algorithm by considering the
  improvement in matching accuracy for poor quality
  prints.
• Compares favorably with Gabor filter based
  enhancement scheme.

25
Thanks for your attention!

• http//www.cubs.buffalo.edu