Digital watermarking techniques for JPEG2000 scalable ...

Electronic & Electrical Engineering.

Communications Research Group.

Digital watermarking techniques for JPEG2000 scalable image coding

Deepayan Bhowmik The University of Sheffield, Sheffield, UK

Contents

Introduction

Digital watermarking: properties and applications

Discrete wavelet transform and its applications

Scalable image coding and its application in multimedia signal processing

Research techniques for image watermarking for JPEG2000 content adaptation

Recent developments and open questions in this field

05/12/2012 @ ICISS 2012 © The University of Sheffield, UK

Contents

Introduction







Introduction

• A digital watermark is the copyright or author identification information which is embedded directly in the digital media in such a way that it is imperceptible, robust and secure.


Audio watermarking

Image watermarking

Video watermarking


+ =



Example


Example


Example


• Imperceptibility

Measurement metrics: RMSE, PSNR, SSIM etc.


Digital watermarking : properties



• Robustness

Measurement metric: Hamming distance, Correlation etc.





• Robustness

Measurement metric: Hamming distance, Correlation etc.

• Fragility

• Tamper-resistance

• Data payload



Applications

• Copyright protection.

• Owner identification.

• Content authentication.

• Broadcast monitoring.

• Transaction tracking.

• Media digital rights management (DRM) in content supply chain.

• Tamper proofing.


Contents

Introduction







Discrete Wavelet Transform (DWT) and applications


DWT


2

2

H

L

2

2

HH

HL

2

2

LH

LL

2D wavelet decomposition

G0

G1

G0

G0

G1

G1

I

DWT


HH1

HL1

LH1

HH2 LH2

HL2 LL2

2 level decomposition

DWT Applications

• Compression

• JPEG-2000 image compression

• MC-EZBC video compression

• De-noising

• Edge-detection

• Image retrieval

• Gait analysis

• Digital communication and many others


Contents

Introduction







Scalable image coding



JPEG2000 image coding using DWT


JPEG2000


JPEG2000 image coding




0

1

0

1

Bit plane 0

Bit plane N

Bit plane N-1

Bit plane N-2

Most

significant

Least

significant LH2

HL2

HH2

LH1

HL1

HH1

HH2

HH1

LH2

LH1

HL2 HL1

LL2

LL2



LL1 HL1

LH1 HH1

HH2

HL2

LH2

LL2

Quarter Resolution

Half Resolution

Full Resolution

Resolution scalability






Quality scalability


Quality scalability


Quality scalability


Example


Contents

Introduction







Robust watermarking techniques for scalable coding


Watermarking techniques


()Embedding

Original Image

Watermarked Image

)(I

)'(IWatermark )(W

()Extraction Test Image Extracted

watermark )'(I

)'(W

Original Image

(for non-blind type) )(I Original watermark )(W

Watermark

detection

decision

Authenti- cation



Embedding Domain

Pixel Domain

Transform Domain

Fourier DCT Wavelet

Non-Blind Blind



Host

Image Watermarked

Image

Authentication

Decision

DWT Coefficient

selection

and watermark

embedding

IDWT Attack including

content

adaptation

DWT

Test

image

Watermark

Extraction

(Blind / Non-blind)

Extracted

Watermark Comparison of

original and

extracted watermark

Watermark



Embedding

where C’m,n = modified coefficient, Cm,n = wavelet coefficient to be modified, wm,n = watermark information, α = watermark strength parameter, wext = extracted watermark.

Extract

,,,,,' nmnm wCCC nmnm ,'

C

CCextw

05/12/2012 @ IET/IPR 2012 © The University of Sheffield, UK

Wavelet based image compression (Bit-plane discarding):

,

Q

CCq

Nk 2).1( Nk 2.Nk 2).1(

2

122).1()1(

NN

k kC2

122.

NN

k kC

C

,2

1.ˆ

QCQC q

where Cq = quantised coefficient, C = original coefficient, Q = quantisation parameter,

Cˆ = decoded coefficient Q = 2N

k € ±1, ±2, ±3… Ck = Center point value


Embed 1: C & C’ in different cluster:

Algorithm 1:

Nk 2).1( Nk 2.Nk 2).1(

)1( kCkC

11

2.

w

k N

C

'C

.2.1

2.

1

NN

kCw

k

w1 = value to embed ‘1’.


Robustness Analysis :

)1( kCkC

T

Ck

1

Nk 2).1( Nk 2.Nk 2).1(

C

'C

Embed 1: C & C’ in same cluster:

.1

2.T

CCk kN

T = Threshold for watermark detection.



)1( kCkC

11

2.

w

k N

T

Ck

1

Nk 2).1( Nk 2.Nk 2).1(

C

'C

Embed 1: combined:

.11

2.

1 T

CC

w

k kN



)1( kCkC

01

2.

w

k N

Nk 2).1( Nk 2.Nk 2).1(

C

'C


.2.1

2.

0

NN

kCw

k

w0 = value to embed ‘0’.



)1( kC kC

01

2.

w

k N

T

C k

1

)1(

Nk 2).1( Nk 2.Nk 2).1(

C

'C


.1

2.

1 0

)1(

w

kC

T

C Nk



Nk 2).1( Nk 2.Nk 2).1(

)1( kCkC

11

2.

w

k N

T

Ck

101

2.

w

k N

T

C k

1

)1(

C

Embed 1 or 0 for correct watermark extraction:

.1

2.

1

2.

01 w

kC

w

k NN


Simulation results :

Map of original coefficient values (C) to retain the watermark information.


Effect of bit-plane based coefficient selection procedure against JPEG2000 quality scaling considering 64:1 compression ratio. Quantisation steps: Q = 20 to Q = 29.

0 1 2 3 4 5 6 7 8 90.1

0.15

0.2

0.25

0.3

0.35

Effect of coefficient selection on JPEG2000 compression for different N. CR = 64:1

N

Ha

mm

ing

Dis

tan

ce

Image 1

Image 2

Image 3

Image 4

Image 5

Image 6

Simulation results :


Wavelet based image compression (Bit-plane discarding):

,

Q

CCq

Nk 2).1( Nk 2.Nk 2).1(

2

122).1()1(

NN

k kC2

122.

NN

k kC

C

,2

1.ˆ

QCQC q

where Cq = quantised coefficient, C = original coefficient, Q = quantisation parameter,

Cˆ = decoded coefficient Q = 2N

k € ±1, ±2, ±3… Ck = Center point value


0 2 3 )(n )1( n

C

0

0 01 1

2)%1( n 2%n

1

2/)(2 n 2/)1(2 n2/)12( n

)1( n

C

0 1

4/)(4 n 4/)12(2 n4/)14( n

C

0 1C

8/)14(2 n 8/)12(4 n8/)38( n

Algorithm 2:

M2


0 1

12% n

0 1

0 1

0 1

0 1

Tree for C: 1011011)2%()( nCb

Tree depth d

Embedding algorithm

....,3,2,1,0,2%

2

iC

bi

i

142C 3225

01000)( Cb 6d


Tree based watermarking rule

Symbols Binary Tree Watermark association

Embedded Zero (EZ) 000xxxx 0

Embedded Zero (EZ) 001xxxx 0

Cumulative Zero (CZ) 010xxxx 0

Weak One (WO) 011xxxx 1

Weak Zero (WZ) 100xxxx 0

Cumulative One (CO) 101xxxx 1

Embedded One (EO) 110xxxx 1

Embedded One (EO) 111xxxx 1

Experimental Results

Existing Algorithm Proposed Algorithm

PSNR Data capacity (L)

PSNR Data capacity (L)

Boat (704x576)

86.40 53.74 2112 84.13 47.43 6336

Barbara (704x576)

80.64 55.12 2112 81.71 49.13 6336

Blackboard (704x576)

69.12 56.45 2112 69.12 50.51 6336

Light House (768x512)

84.48 55.36 2048 82.43 48.78 6144



0 5 10 15 20 25 30 35 40 45 50

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

JPEG2000 compression Ratio

Ha

mm

ing

Dis

tan

ce

Robustness against JPEG2000: Blackboard

d=5 (=31)

d=5 (=329)

d=6 (=112)

d=6 (=1278)

d=7 (=466)

d=7 (=4980)

0 1 2 3 4 5 6 7

0

0.1

0.2

0.3

0.4

0.5

p

Ha

mm

ing

Dis

tan

ce

Robustness against bit plane discarding: Blackboard

d=5 (=31)

d=5 (=329)

d=6 (=112)

d=6 (=1278)

d=7 (=466)

d=7 (=4980)



0 5 10 15 20 25 30 35 40 45 50

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4


Ha

mm

ing

Dis

tan

ce

Robustness against JPEG2000: Light House

d=5 (=31)

d=5 (=335)

d=6 (=125)

d=6 (=1358)

d=7 (=482)

d=7 (=5129)

0 1 2 3 4 5 6 7

0

0.1

0.2

0.3

0.4

0.5

p

Ha

mm

ing

Dis

tan

ce

Robustness against bit plane discarding: Light House

d=5 (=31)

d=5 (=335)

d=6 (=125)

d=6 (=1358)

d=7 (=482)

d=7 (=5129)



0 10 20 30 40 50 600

0.05

0.1

0.15

0.2

0.25


Ha

mm

ing

Dis

tan

ce


Proposed algorithm

Existing algorithm

0 10 20 30 40 50 600.05

0.1

0.15

0.2

0.25

0.3


Ha

mm

ing

Dis

tan

ce


Proposed algorithm

Existing algorithm



0 10 20 30 40 50 600

0.05

0.1

0.15

0.2

0.25


Ha

mm

ing

Dis

tan

ce


Proposed algorithm

Existing algorithm

0 10 20 30 40 50 600.05

0.1

0.15

0.2

0.25

0.3


Ha

mm

ing

Dis

tan

ce


Proposed algorithm

Existing algorithm

Contents

Introduction







Demo

• Watermarking demo.



Open questions in this field and comments

• Watermarking a growing research area for digital copyright protection and other applications.

• Scalable coded contents are being used for various applications including mobiles.

• Watermarking for scalable coded contents are challenging and needs more research.

• Video watermarking is a promising research area to explore.

Pointers

• WEBCAM Framework developed by The University of Sheffield.

• http://svc.group.shef.ac.uk/webcam.html


Pointers


Questions and Answers:



Thank You All

[email protected]

www.bhowmik.net/publication.html



Digital watermarking techniques for JPEG2000 scalable ...

Documents