Use of Exterior Contours and Shape Features in Off-line Signature - - PowerPoint PPT Presentation
Use of Exterior Contours and Shape Features in Off-line Signature Verification Siyuan Chen and Sargur Srihari Center of Excellence for Document Analysis and Recognition (CEDAR) University at Buffalo State University of New York, USA Overview
Known Signatures Questioned Signature
Questioned is Genuine/Forgery/Unknown
Verification process
Overview of algorithm Combining contours of signature Matching contours of signature Feature extraction
Grey scale image Binary image
before after
Determine threshold Threshold
Image Preprocessing
(binarization,repair)
Contour Generation
(chain code, pseudo path)
Matching to reference contour by DTW Feature Extraction
(Zernike moments of contour segs)
Reference contour Randomly select one as reference Compute distance with known set of n images
,
33.42, 53.94, 35.30 66.55, 13.62, 73.84 17.30, 13.58 … … … … … …
Genuine/Forgery/Unknown
20 640
Exterior Contours
(1) (2) (3) (4) (5)
Contour (5) X: 365 365 365 … Y: 96 95 94 … Slope: 2 2 2 … Curvature: 7 0 0 … Contour (2) X: 68 68 68 … Y: 91 90 89 … Slope: 2 2 2 … Curvature: 7 0 0 … Contour (1) X: 9 10 11 … Y: 104 104 104 … Slope: 2 2 3 … Curvature: 0 0 1 … Contour (3) X: 297 298 299 … Y: 53 52 51 … Slope: 3 3 3 … Curvature: 0 0 0 … Contour (4) X: 351 352 353 … Y: 108 107 106 … Slope: 2 3 3 … Curvature: 7 1 0 …
direction direction
1 2 2
)) ( ), ( ( )) ( ) ( ( ) , ( here w ), , ( ) , (
y x c y x s y x
i curvature i curvature f i slope i slope f i i d d D + − = =
∑ − × − − = + =
=
x
L l y x y x y x y x y x y x y i x i y x
l T m l T l T d i i i i i i i i i i D i i D
' ' ' ' ' ' ' ' , '
) ( )) ( ), ( ( )) , ( ), , ' (( )) , ( ), , ' (( ) , ( min ) , ( φ φ ξ ξ
y x y x A
T T T T D Y X d + − − = ) 1 , 1 ( ) , (
X Y
x y
θ
ρ
Order Zernike Moments No. A00 1 1 A11 1 2 A20, A22 2 3 A31,A33 2 4 A40,A42, A44 3 5 A51, A53, A55, 3 6 A60, A62, A64, A66, 4
16
32
1 ) , ( ) , ( 1 )! 2 ( )! 2 ( ! )! ( ) 1 ( ) ( ) ( ) , ( ) , (
2 2 * 2 2 / ) (
≤ + ∑ ∑ + = ∑ − − − + − − = = =
− − =
y x V y x f n A s m n s m n s s n R e R V y x V
nm x y nm s n m n s s nm jm nm nm nm
θ ρ π ρ ρ ρ θ ρ
θ
⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦ ⎤ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ ⎡ L L 18 . 12 68 . 6 59 . 98 77 . 32 49 . 86 29 . 54 84 . 67 42 . 40
a 32 length feature vector qi a 32 length feature vector ki
⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦ ⎤ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ ⎡ L L 58 . 13 30 . 17 84 . 73 62 . 13 55 . 66 30 . 35 94 . 53 42 . 33
Ki : Qi :
( )
∑ − = ∑ =
= = 32 1 2 20 1
signatures
segment ith e between th distance Euclidean the is Where 1 1 ) , ( : signatures known and questioned between distance Harmonic
j ij ij i i i i
k q d d d K Q D
Gradient (12 bits): 111101111111 Structural (12 bits): 000011001100 Concavity (8 bits): 10100000 8 4
s "1" have rs both vecto where bits
number the is s "0" have rs both vecto where bits
number the is 1024 5 . score Similarity bits 1024 8 4 8 12 12 bits Total
11 00 11 00
C C C C + × = = × × + + =
Questioned Knowns
Shape Algorithm Contour Algorithm
Sgsc Szer Threshold set 1 Threshold set 2
high confidence genuine / high confidence forgery / low confidence genuine / low confidence forgery
high confidence genuine vs. low confidence forgery or high confidence genuine vs. high confidence genuine or high confidence genuine vs. low confidence genuine
Genuine
high confidence forgery vs. low confidence genuine or high confidence forgery vs. high confidence forgery or high confidence forgery vs. high confidence forgery
Forgery
high confidence genuine vs. high confidence forgery
Unknown
Genuine signatures (1320): 55 individuals , 24 signatures each Forgery signatures (1320): 55 individuals , 24 signatures each
(with rejection)
(with rejection)
(with rejection)
Rejection Rate Accuracy