View
355
Download
3
Category
Preview:
DESCRIPTION
T.Q. Pham, S.W. Perry and P.A. Fletcher, DICTA, 2009.
Citation preview
Copyright CISRA Slide 1 Printed 30 November 2009
Australian R&D with global impact
Paper fingerprinting using alpha-masked image matching
Tuan Pham, Stuart Perry, and Peter FletcherCanon Information Systems Research Australia
Paper 55 Session3B 12:10PM-12:30PM on Thu 3 Dec
Copyright CISRA Slide 2 Printed 30 November 2009
Australian R&D with global impact
Overview
1. Paper FingerPrint (PFP)Random structure of paperUniqueness
2. Alpha-masked image matchingNormalized correlationImage inpainting
3. PFP robustnessEvaluationPossible improvements
Copyright CISRA Slide 3 Printed 30 November 2009
Australian R&D with global impact
Paper fingerprintIntrinsic characteristic of a piece of paper that uniquely describes itself
The coupon
Paper Fingerprint SystemPaper Fingerprint System
The original fingerprint
Match Strength
ThresholdComparison
Not Original Original
Document authentication application
Copyright CISRA Slide 4 Printed 30 November 2009
Australian R&D with global impact
Paper FingerPrinting (PFP) using a scanner
Officepaper
Copyright CISRA Slide 5 Printed 30 November 2009
Australian R&D with global impact
PFP matching using cross-correlationPaper FingerPrint is a 256x256 8-bit grayscale scan at 600dpi
PFP match is determined based on correlation peak strength:Peak strength > 10 → matchPeak strength < 10 → non-match
Using extreme value theory, the false alarm rate is about 1x10-50
___ Fisher-Tippettdistribution fit
Non-Match PFP strengths Matching PFP strengths
Copyright CISRA Slide 6 Printed 30 November 2009
Australian R&D with global impact
Correlation is not robust against changeWe swap 7.5% of pixels around → match strength drops below 10
Printing also decreases match strength & increases false matches
Some printed textcreates a false negative
Low correlation due to the influence of the printed
areas
Same sheet of paper
Same printed textcreates a false positive
Same printed textcreates a false positive
High correlation due to the influence of the
printed areas
Different sheets of paper
Copyright CISRA Slide 7 Printed 30 November 2009
Australian R&D with global impact
Solution 1: alpha-masked correlation [Fitch et al]
Use weight to mask out change areas during least-square matching:
( )20 0 1 2 0 0 1 2 0 0
,
2 21 1 2 1 1 2 2 1 2 2
( , ) ( , ) ( , ) ( , ) ( , )
2x y
E x y f x y f x x y y x y x x y y
f f f f
α α
α α α α α α
= − − − − −
= ⊗ − ⊗ + ⊗
∑α
1 1 2 2
1 2
2Nf fE α αα α⊗
= −⊗
If images f1 & f2 have zero mean, alpha-masked correlation can be simplified to normalized correlation:
α
Alpha-masked correlation
Peak strength = 4.04
Paper 1
Same printed texts
Paper 2
Cross-correlation
Peak strength = 12.44
Copyright CISRA Slide 8 Printed 30 November 2009
Australian R&D with global impact
Solution 2: inpainting followed by correlation
How well do these different solutions compare to each other?
Scanned paper with printed texts Filled-in with mean value Smooth inpainting
Normalized correlation is most discriminative, followed by mean-filled correlation
5.287.597.635.67Ratio
5.365.885.848.16Non-match
28.3444.6244.6446.32Match
InpaintingMean-filledNormalizedAlpha-masked
Copyright CISRA Slide 9 Printed 30 November 2009
Australian R&D with global impact
Fill factor experiment: synthetic maskMask is successively eroded to reduce the fill-factor
…
fill = 0.77
fill = 0.58
fill = 0.45
fill = 0.05
PFP 1
Paper 1 after printing
PFP 2
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.80
5
10
15
20
25
30
35
40
45
50
mask fill factor
mat
ch s
treng
th
alpha-masked correlation [8]normalized correlation (section 2.3)mean-filled correlation (section 3.3)inpainting correlation (section 3.3)correlation of non-matching pairs
100
90
80
70
60
50
40
30
20
10
0
Copyright CISRA Slide 10 Printed 30 November 2009
Australian R&D with global impact
The document is scanned twice, 256x256 image patches are matched
Match strength vs. fill factor from a real document
…
fill = 0.94
fill = 0.86
fill = 0.73
fill = 0.16 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10
5
10
15
20
25
mask fill factor
mat
ch s
treng
th
alpha-masked correlation [8]
50
40
30
20
10
0
Copyright CISRA Slide 11 Printed 30 November 2009
Australian R&D with global impact
Improve PFP by multiple orientation scans
Reflection from paper consists of diffuse and specular components
These components can be separated by photometric stereography
_+++
paper scanned at 0º paper scanned at 180º
diffuse reflectance specular reflectance
Copyright CISRA Slide 12 Printed 30 November 2009
Australian R&D with global impact
Improve PFP robustness by double-sided scan
Verify the Paper FingerPrint on both sides.
Front and back side must correlate
The displacement between front and back side is fixed
Front side scanned at 0°
Back side scanned at 0°
Correlation of front and back
Copyright CISRA Slide 13 Printed 30 November 2009
Australian R&D with global impact
ConclusionsConventional scanners capture internal structure of paper: Paper FingerPrint (PFP)
PFPs are very unique and can be used for authentication
PFPs can be matched even after printing
PFPs can be made more robust using more than one scans
Copyright CISRA Slide 14 Printed 30 November 2009
Australian R&D with global impact
Thank you
tuan.pham@cisra.canon.com.au
Recommended