![Page 1: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/1.jpg)
Image Processing forBiometric ApplicationsRuggero Donida Labati
Image Processing Techniques forDifferent Biometric Traits
Academic year 2016/2017
![Page 2: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/2.jpg)
• Fingerprint• Iris• Face
Summary
![Page 3: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/3.jpg)
Fingerprint
![Page 4: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/4.jpg)
• A fingerprint in its narrow sense is an impression left by the friction ridges of a human finger
• One of the most used traits in biometric applicationso High durabilityo High distinctivity
• The more mature biometric trait
Fingerprint
Ridge Valley
![Page 5: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/5.jpg)
• Level 1: the overall global ridge flow pattern
• Level 2: points of the ridges, called minutiae points
• Level 3: ultra-thin details, such as pores and local peculiarities of the ridge edges
Levels of analysis
![Page 6: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/6.jpg)
• Usually from 0°to 180°• Evaluation of the gradient orientation in
squared regionsFor each local region, the ridge orientation is considered as the mean of the angular values of the pixels
Level 1: local ridge orientation
![Page 7: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/7.jpg)
• Global or map
• x-signature
Level 1: local ridge frequency
![Page 8: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/8.jpg)
Level 1: singular regions
Loop Delta Whorl
• Pointcaré algorithm
• The northern loop is usually considered as the core point
![Page 9: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/9.jpg)
Level 1: Pointcaré algorithm
• Let G is be a vector field and C be a curve immersed in G, then the Poicaré index PG,C is defined as the total rotation of the vectors of G along C
• Considering eight near elements dk (k = 0… 7)
.
• Classification– 0° = the point is not a singular point;– 360° = whorl region;– 180° = core point;– -180° = delta point.
![Page 10: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/10.jpg)
Level 1: ridge count
![Page 11: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/11.jpg)
• Evaluates both frequency and space
Level 1: Gabor filters (1/2)
![Page 12: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/12.jpg)
The even-symmetric Gabor filter has the general form
where Φ is the orientation of the Gabor filter, f is the frequency of a sinusoidal plane wave, and σxand σy are the space constants of the Gaussian envelope along x and y axes, respectively.
Level 1: Gabor filters (2/2)
![Page 13: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/13.jpg)
Level 2: minutiae
termination bifurcation lake point orisland
Independentridge
spur crossover
![Page 14: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/14.jpg)
Level 2: the most used minutiae extractor
Crossing number: 8 neigborhood of the pixel p
𝐶𝐶𝐶𝐶 𝑝𝑝 =12�𝑘𝑘=1
8
𝑛𝑛𝑘𝑘 − 𝑛𝑛 𝑘𝑘+1 mod 8
If 𝐶𝐶𝐶𝐶 𝑝𝑝 = 1 or 𝐶𝐶𝐶𝐶 𝑝𝑝 = 3, p is a minutia
![Page 15: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/15.jpg)
Level 2: minutia patterns
![Page 16: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/16.jpg)
Level 2: ridge following
![Page 17: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/17.jpg)
Level 3
• Minimum resolution of 800 DPI
![Page 18: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/18.jpg)
The biometric recognition process
![Page 19: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/19.jpg)
Acquisition
Latent Inked and rolled Live scan
![Page 20: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/20.jpg)
Thermal (Sweeping)
3dimensional
Images
![Page 21: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/21.jpg)
Quality evaluation (1/2)
NIST NFIQ
![Page 22: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/22.jpg)
• Local standard deviation (16 X 16 pixels)
Preprocessing: segmentation
![Page 23: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/23.jpg)
Preprocessing: enhancement
• Contextual filtering
![Page 24: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/24.jpg)
Minutiae-based matching
• Two minutiae are considered as correspondent if their spatial distance sd and direction difference dd are less than fixed thresholds
• Non-exact point pattern matchingo Roto-translationso Non-linear distortionso False minutiaeo Missed minutiaeo Non constant number of minutiae
![Page 25: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/25.jpg)
Minutiae-based matching:Delaunay triangulation
• Featureso Side lengths of th triangles
o Greater angls
o Incenters
o Minutiae features (x, y, θ)
![Page 26: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/26.jpg)
Minutiae-based matching:Delaunay triangulation (1/2)
• Method A o Method B
![Page 27: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/27.jpg)
Minutiae-based matching:NIST Bozhorth 3
• Step 1: Construction of Intra-Fingerprint Minutiae Comparison Tables
• Step 2: Construction of Inter-Fingerprint Compatibility Table Step 3: Traverse Inter-Fingerprint Compatibility Table constructed in second step
![Page 28: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/28.jpg)
Other minutiae-based recognition methods: Cylindercode
![Page 29: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/29.jpg)
[1] A. K. Jain, et al., “Filterbank-based fingerprint matching,”IEEE Transactions on Image Processing, 2000.
Template Finercode
Input Finercode
Euclidean distance Matchingresult
Compute A.A.D.featureFiltering
Normalizeeach sectorInput image
Divide imagein sectors
Locate thereference point
Other recognition methods: Fingercode
![Page 30: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/30.jpg)
Iris
![Page 31: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/31.jpg)
The colored ring around the pupil of the eye is called the Iris
Iris biometric trait
![Page 32: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/32.jpg)
Iris-based identification
AcquisitionModule
Feature Extraction
ModuleDataBase
iriscode
MatchingModule
N iriscode
Identity / Not identified
AcquisitionModule
Feature Extraction
Module
QualityChecker
Enrollment
Identification
QualityChecker
iriscode
![Page 33: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/33.jpg)
Iris scanners
![Page 34: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/34.jpg)
How iris recognition works
![Page 35: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/35.jpg)
Iris segmentation using Hough transform
![Page 36: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/36.jpg)
Iris segmentation using Daugman’s method
![Page 37: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/37.jpg)
Iris segmentation using active contours
![Page 38: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/38.jpg)
I
III = separable eyelashes
II = not-separable eyelashes
Eyelids and eyelashes
![Page 39: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/39.jpg)
Separable eyelashes
GaborFilter
Thresholding
𝑔𝑔(𝑥𝑥, 𝑦𝑦) = 𝐾𝐾 exp� −𝜋𝜋(𝑎𝑎2(𝑥𝑥 − 𝑥𝑥0)2 + 𝑏𝑏2(𝑦𝑦 − 𝑦𝑦0)2)� exp�𝑗𝑗(2𝜋𝜋𝐹𝐹0 (𝑥𝑥 cos𝜔𝜔0 + 𝑦𝑦 sin𝜔𝜔0) + 𝑃𝑃)�
𝐿𝐿1 (𝑥𝑥, 𝑦𝑦) = �1 if �𝐼𝐼(𝑥𝑥, 𝑦𝑦) ∗ ℛℯ�𝑔𝑔(𝑥𝑥, 𝑦𝑦)�� > 𝑡𝑡3 0 else
�
𝑡𝑡3 = 𝑘𝑘 max(|𝐼𝐼𝑟𝑟(𝑥𝑥, 𝑦𝑦) ∗ 𝑔𝑔(𝑥𝑥, 𝑦𝑦)|)
![Page 40: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/40.jpg)
Non-separable eyelashes and reflections
1)
2)
3)
4)5) Starting from an initial set of points L5, we
apply an iterative region growingtechnique based on the following
where μ and σ are the local mean andstandard deviation of I(x,y) processed in a3 × 3 matrix centered in the point x,yand β is a fixed threshold .
𝐿𝐿2 (𝑥𝑥, 𝑦𝑦) = �1 if var�𝐼𝐼(𝑥𝑥, 𝑦𝑦)� > 𝑡𝑡4 0 else
�
𝐿𝐿3 (𝑥𝑥, 𝑦𝑦) = erode(𝐿𝐿2(𝑥𝑥, 𝑦𝑦), 𝑆𝑆 )
𝐿𝐿4 (𝑥𝑥, 𝑦𝑦) = �1 if 𝐼𝐼(𝑥𝑥, 𝑦𝑦) > 𝑡𝑡50 else
�
𝐿𝐿5 = 𝐿𝐿3 OR 𝐿𝐿4
𝜇𝜇 + 𝛽𝛽𝜎𝜎 < 𝐼𝐼(𝑥𝑥, 𝑦𝑦)
![Page 41: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/41.jpg)
Iris normalization: motivation
Pupil dilation Gaze deviation Resolution
![Page 42: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/42.jpg)
Iris normalization: Rubber sheet model
![Page 43: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/43.jpg)
Iriscode (1/2)
”The detailed iris pattern is encoded into a 256-byte “IrisCode” by demodulating it
with 2D Gabor wavelets, which represent the texture by phasors in the
complex plane. Each phasor angle (Figure*) is quantized into just the
quadrant in which it lies for each local element of the iris pattern, and this
operation is repeated all across the iris, at many different scales of analysis”
J. Dougman
Parte reale di h
Parte immaginaria h
11
(*)
Real part of h
Imaginary part of h
![Page 44: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/44.jpg)
Iriscode (2/2)
![Page 45: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/45.jpg)
Matching Score = 0.26848
Iriscode matching
![Page 46: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/46.jpg)
Face
![Page 47: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/47.jpg)
Face recognition using still images
![Page 48: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/48.jpg)
Face acquisition sensors
• Cameras (a)• Webcams• Scanners for off-line
acquisitions• Termocameras (b)• Multispectral devices
(c,d,e,f)
![Page 49: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/49.jpg)
• Scan window over image
• Classify window as either:– Face– Non-face
ClassifierWindowFace
Non-face
Face detection
![Page 50: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/50.jpg)
• Consider an n-pixel image to be a point in an n-dimensional space, x Rn.
• Each pixel value is a coordinate of x.∈
x1
x2
x3
Images as features
![Page 51: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/51.jpg)
{ Rj } are set of training images.
x1
x2
x3R1 R2
I
),(minarg IRdistID jj
=
Nearest Neighbor Classifier
![Page 52: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/52.jpg)
• Use Principle Component Analysis (PCA) to reduce the dimsionality
Eigenfaces
![Page 53: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/53.jpg)
Eigenfaces (2)
[ ] [ ][ x1 x2 x3 x4 x5 ] W
• Construct data matrix by stacking vectorized images and then apply Singular Value Decomposition (SVD)
![Page 54: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/54.jpg)
• Modelingo Given a collection of n labeled training images
Compute mean image and covariance matrix Compute k Eigenvectors (note that these are images) of covariance
matrix corresponding to k largest Eigenvalues Project the training images to the k-dimensional Eigenspace
• Recognitiono Given a test image, project to Eigenspace
Perform classification to the projected training images
Eigenfaces (2)
![Page 55: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/55.jpg)
Problems of PCA
![Page 56: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/56.jpg)
Fisherfaces
• An n-pixel image x∈Rn can be projected to a low-dimensional feature space y∈Rm by
y = Wx
where W is an n by m matrix.
• Recognition is performed using nearest neighbor in Rm.
• How do we choose a good W?
![Page 57: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/57.jpg)
• PCA (Eigenfaces)
Maximizes projected total scatter
• Fisher’s Linear Discriminant
Maximizes ratio of projected between-class to projected within-class scatter
WSWW TT
WPCA maxarg=
WSW
WSWW
WT
BT
Wfld maxarg=
χ1χ2
PCA
LDA
PCA and LDA
![Page 58: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/58.jpg)
Face recognition based on fiducial points
![Page 59: Image Processing Techniques for Different Biometric Traitshomes.di.unimi.it/donida/biometricsphd/Lecture2_Donida...Image Processing for Biometric Applications Ruggero Donida Labati](https://reader033.vdocument.in/reader033/viewer/2022050422/5f91198abf0dfc0e1d1bd5a0/html5/thumbnails/59.jpg)
Gabor wavelet Discrete cosine transform
Examples of other features