Robust Reversible Watermarking

chemistoddAI and Robotics

Nov 6, 2013 (4 years and 2 days ago)

83 views

Robust Reversible Watermarking
via Clustering and

Enhanced Pixel
-
Wise Masking

Source
:

IEEE Transactions on Image Processing,
Vol. 21, No. 8, pp. 3598
-

3611, August 2012

Authors
:
Lingling

An,
Xinbo

Gao
,
Xuelong

Li,
Dacheng

Tao, Cheng Deng
and
Jie

Li

Presenter
: Huynh Ngoc Tu

Date
:
October 18
th

,
2012

OUTLINE


INTRODUCTION


RELATED WORK


PROPOSED SCHEME


EXPERIMENTAL RESULTS


CONCLUSIONS


2

INTRODUCTION


Issues of robust reversible watermarking mechanisms:


Reversibility: how to handle both overflow and underflow of pixels


Robustness: how to resist unintentional attacks


Invisibility: how to make a trade
-
off between robustness and invisibility

3

INTRODUCTION


Features of the proposed scheme:


Offer the robust and lossless watermark embedding


Outperform in term of reversibility, robustness, invisibility, capacity
and run
-
time complexity


Be applicable to different kinds of images


Be readily applicable in practice




4

5


PWM: pixel
-
wise
marking proposed by
Barni

et al.


Revealing the perceptual characteristics of HVS


Compute the
JND

threshold of each wavelet coefficient based on resolution sensitivity,
brightness sensitivity, and texture sensitivity


Estimate
how HVS perceives disturbances in images

Related Works


Pixel
-
wise marking (PWM)

Related Works


Pixel
-
wise marking (PWM)

6

DWT
A
f
3
18
20
-
9
8
30
15
3
-
4
6
1
7
-
4
5
-
8
-
6
-
2
A
D
V
H
LL

H
L

LH

H
H

(
)
(
)
(
)
(
)
0.02
,,,,,,
JND i j i j p i j
w
r
r w r
= Q Y P
(
)
(
)
: resolution
1.00, if 0
0.32, if 1
2, if
,.
0.16, if 2
1, otherwise
0.10, i
,
f 3
HH
r
r
w
r w
r
r
r
w
ì ü
ï ï
=
ï ï
ï ï
ì ü
ï ï
ï ï
=
ï ï
=
ï ï
ï ï ï ï
Q =
í ý í ý
ï ï ï ï
=
ï ï ï ï
ï ï
î þ
ï ï
ï ï
=
ï
ï
î þ
Q
ï
ï
(
)
(
)
(
)
(
)
(
)
(
)
3
3 3
: brightness sensitivity
1,, if ,,0.5
,,1
,,, otherwise
1
where ,,1,1
256
,
2
,
2
LL
i j i j
i j
i j
i j
i j c
i j
r
r
r
r
r
r
r r
ì ü
ï ï
- L L <
ï ï
ï ï
Y = +
í ý
ï ï
L
ï ï
ï ï
î þ
æ ö
ê ú ê ú
÷
ç
ê ú ê ú
÷
ç
L = + +
÷
ç
ê ú ê ú
÷
ç
÷
- -
è ø
ë û
Y
ë û
7

DWT
A
f
3
18
20
-
9
8
30
15
3
-
4
6
1
7
-
4
5
-
8
-
6
-
2
A
D
V
H
LL

H
L

LH

H
H

(
)
(
)
(
)
(
)
0.02
,,,,,,
JND i j i j i j
w
r
r w r r
= Q Y P
(
)
(
)
(
)
3
1 1
0 0 0
,,
3
3 3
: texture sensitivity
1
,,,
16 2 2

,
1,1
2 2
,
k
k k k
k x y
LH HL HH
LL
i j
i j c x y
i j
Var c
i
x y
j
r
w
r
w
r r
r
r
-
+
= = =
Î
- -
é ù
æ ö
÷
ç
ê ú
÷
Y = + +
ç
÷
ê ú
ç
÷
ç
è ø
ê ú
ë û
ì ü
æ ö
ï ï
÷
ï
ç
÷
´ ³ + + + +
ç
í ý
÷
ç
÷
ï
ç
è
ï
P
ø
î
å å å å
0,1
0,1
x
y
=
=
ï
ï
ï
þ
Proposed Scheme (1/8)

8

PIPA

Module

SQH

Construction

EPWM

Embedding

Host Image

Watermarked
Image

Proposed Scheme (2/8)


Properly inspired pixel adjustment (PIPA)

9

Proposed Scheme (3/8)


Properly inspired pixel adjustment (PIPA)

10

T
he
effects of changing wavelet coefficients on pixels

based on multiple sub
-
bands and multiple wavelet coefficients.

(
)
(
)
(
)
(
)
(
)
'
, if ,2 1
,
, if ,
:adjustment scale
t
I i j I i j
I i j
I i j I i j
h h
h h
h
ì
ï
- > - -
ï
ï
=
í
ï
+ <
ï
ï
î
,2 1
: -bit host image
t
t t
h h
é ù
- -
ê ú
ë û
Pixels are adjusted into a new
range:

Proposed Scheme(4/8)

11

SQH
Construction


Focus on the mean of wavelet coefficients (MWC) histogram


Design in high
-
pass sub
-
band of wavelet decomposition (HVS is less
sensitive)


Proposed Scheme(5/8)

12

SQH
Construction

DWT
A
f
3
18
20
-
9
8
30
15
3
-
4
6
1
7
-
4
5
-
8
-
6
-
2
A
D
V
H
Host Image

Block of HL
coefficients

(
)
(
)
(
)
(
)
1 1
2 2
1
,
2 2
1 26
15 3 1 7
2 2 4
h w
k k
i j
S P i j
h w
- -
= =
=
- -
= + + + =
´
å å
(
)
{
}

,:2 peak points
: a predefined constant for threshold co
nt
,
rol
l r
k
x x
x S
x
d
d
d
Î
£
Compute the Euclidean distance of:

(
)
(
)
(
)
(
)
{
}
If max:
all of th blocks will be retaind for emb
eddi
,
ng
min,,max
l r
d x S d x S
d
³
Þ
Proposed Scheme (6/8)

13

EPWM
-
based Embedding

0 1: turning parameter corresponding to t
he weight of brightness sensitivity
J
£ £
Embed watermark bit
b
k

to block
k
:

(
)
(
)
{
}
(
)
* *
*
,

: obtained MWC after th th watermark bit
is embedded
/
arg min,
l r
w
w
k k k
k
k
x x x
S k
S S abs
S S
S S
S
k
b
d S x
b
bl
Î
= -
=
=
-
+
: watermark strength
: global paramter
l
a
Proposed Scheme (7/8)

14

Extraction based on k
-
means clustering

WMC histogram in a watermarked
image

Bit “1” and “0”

Proposed scheme (8/8)

15

Extraction based on k
-
means clustering

Watermarked
Image

DWT

LH

LL

HL

HH

1
,...,,...,
w w w w
k m
S S S S
é ù
=
ê ú
ë û
Extract:

0, if S Class II
1, if S Class I or Class III
W
k
r
W
k
k
b
ì
ï
Î
ï
ï
=
í
ï
Î
ï
ï
î
Recover the MWCs:

r w r
k k k
S S b
bl
= -
Experimental Results

16

JND
thresholds of different masking models: (a) original


sub
-
band
, (b) PWM, (c) IPWM, and (d) EPWM.

0
LL
c
Experimental Results

17

I
mages
noised by different masking models:

(
a) PWM
, (
b) IPWM, and (c) EPWM with
ϑ
= 0
.
8.

Experimental Results

18

PSNRs for different tuning parameters.

Experimental Results

19

Pure capacities for different thresholds.

Experimental Results

20

PSNRs for different
thresholds

Experimental Results

21

Robustness against JPEG compression for different thresholds.

Experimental Results

22

Robustness against JPEG2000 compression for different thresholds
.

Experimental Results

23

Robustness against AGN for different thresholds

Experimental Results

24

25

26

27

W
atermarked
images of four methods. (a) Host images.

(b) HR. (c) HDC (1). (d) HDC (2). (e) WSQH
-
SC. The PSNRs (dB) are
given below
the watermarked images

28

CONCLUSIONS


Obtains comprehensive performance in term of reversibility,


robustness, invisibility, capacity and runtime complexity


Adjusts the watermark strength for trade
-
off between robustness
and invisibility


Shortfall
: the paper does not analyze its robustness against all the
known attacks


29

30

[11] C. De
Vleeschouwer
, J.
Delaigle
, and B.
Macq
, “Circular
interpretation of
bijective

transformations in lossless watermarking for media
asset
management
,”
IEEE Trans. Multimedia
, vol. 5, no. 1, pp. 97

105, Mar
.
2003
.

[12] Z. Ni, Y. Shi, N. Ansari, W. Su, Q. Sun, and X. Lin, “Robust
lossless image
data
hiding designed for semi
-
fragile image authentication,”
IEEE Trans
. Circuits Syst.
Video Technol.
, vol. 18, no. 4, pp. 497

509, Apr
. 2008
.

[13] D.
Zou
, Y. Shi, Z. Ni, and W. Su, “A semi
-
fragile lossless
digital watermarking
scheme based on integer wavelet transform,”
IEEE Trans
. Circuits
Syst. Video
Technol.
, vol. 16, no. 10, pp. 1294

1300, Oct. 2006.

[29
] M.
Barni
, F.
Bartolini
, and A.
Piva
, “Improved wavelet
-
based watermarking

through pixel
-
wise masking,” IEEE Trans. Image Process
., vol
. 10, no. 5, pp. 783

791, May 2001.

[31]
G.
Xuan
, Y. Shi, P. Chai, X. Cui, Z. Ni, and X. Tong, “
Optimum histogram
pair
based image lossless data embedding,” in
Proc.
Digit Watermark
.
, 2008, pp. 264

278
.