IMAGE ADAPTIVE WATERMARKING USING WAVELET

TRANSFORM


A THESIS

SUBMITTED IN PARTIAL FULFILLMENT OF THE

REQUIREMENTS FOR THE DEGREE OF


MASTER OF TECHNOLOGY

IN

ELECTRONICS SYSTEM AND COMMUNICATION


By

Ms. T MITA KUMARI












Department of Electrical Engineering
National institute of Technology
Rourkela-769008
2007






IMAGE ADAPTIVE WATERMARKING USING WAVELET
TRANSFORM


A THESIS

SUBMITTED IN PARTIAL FULFILLMENT OF THE

REQUIREMENTS FOR THE DEGREE OF


MASTER OF TECHNOLOGY

IN

ELECTRONICS SYSTEM AND COMMUNICATION


By

Ms. T MITA KUMARI

Under the Guidance of
Dr. SUPRAVA PATNAIK





Department of Electrical Engineering
National institute of Technology
Rourkela-769008
2007







National institute of Technology
Rourkela

CERTIFICATE


This is to certify that the thesis entitled Image Adaptive Watermarking using Wavelet
Transform submitted by Ms. T Mita Kumari, in partial fulfillment of the requirements for the
award of Master of Technology in the Department of Electrical Engineering, with specialization
in Electronics System and Communication at National Institute of Technology, Rourkela
(Deemed University) is an authentic work carried out by her under my supervision and guidance.
To the best of my knowledge, the matter embodied in the thesis has not been
submitted to any other University/Institute for the award of any Degree or Diploma.












Dr. Suprava Patnaik
Asst. Professor
Department of Electrical Engineering
NATIONAL INSTITUTE OF TECHNOLOGY
Rourkela-769008

Date:



ACKNOWLEDGEMENTS

On the submission of my thesis report of Image Adaptive Watermarking Using
Wavelet Transform, I would like to extend my gratitude & my sincere thanks to my supervisor
Dr. Suprava Patnaik, Asst. Professor, Department of Electrical Engineering for her constant
motivation and support during the course of my work in the last one year. I truly appreciate and
value her esteemed guidance and encouragement from the beginning to the end of this thesis. I
am indebted to her for having helped me shape the problem and providing insights towards the
solution.
I express my gratitude to Dr.P.K.Nanda, Professor and Head of the Department,
Electrical Engineering for his invaluable suggestions and constant encouragement all through the
thesis work.
I will be failing in my duty if I do not mention the laboratory staff and administrative
staff of this department for their timely help.
I would like to thank all whose direct and indirect support helped me completing my
thesis in time.
This thesis would have been impossible if not for the perpetual moral support from my
family members, and my friends. I would like to thank them all.


T Mita Kumari
M.Tech (Electronics System and Communication)










i
CONTENTS

ABSTRACT iii
LIST OF FIGURES v
LIST OF ACRONYMS vi
1 INTRODUCTION
1.1 Introduction 1
1.2 Watermarking System 2
1.3 Watermarking Requirements .. 4
1.3.1 Imperceptibility 4
1.3.2 Robustness 5
1.3.3 Capacity 7
1.4 Watermarking Applications 8
1.5 Contribution of the Thesis and Chapter Organization ........................................ 10
2 WATERMARKING ATTACKS AND PERFORMANCE MEASUREMENTS
2.1 Introduction 11
2.2 Classification of Attacks . 11
2.2.1 Malicious Attacks .. 12
2.2.2 Non-Malicious Attacks . 12
2.2.3 Removal Attacks 13
2.2.4 Geometric Attacks . 14
2.2.5 Cryptographic Attacks ... 14
2.2.6 Protocol Attacks . 15
2.3 Performance Measures of Watermarking Algorithms . 16
2.4 Literature Review 17
2.5 Chapter Summary 22
3 FUSION-BASED WATERMARKING
3.1 Introduction 23
3.2 Algorithm Description ... 24
3.2.1 Watermark Embedding Method 24
3.2.2 Watermark Extracting Method . 27
ii
3.3 Simulation results and discussion .. 28
3.4 chapter Summary ... 37
4 SPREAD SPECTRUM-BASED WATERMARKING
4.1 Introduction 38
4.2 Algorithm Description ... 38
4.2.1 Watermark Embedding Method . 38
4.2.2 Watermark Extracting Method .. 41
4.3 Simulation Results and Discussion 41
4.4 Chapter Summary .. 49
5 CONCLUSION
5.1 Conclusion . 50
5.2 Future Work 51
REFERENCES 52














iii
ABSTRACT

The availability of versatile multimedia processing software and the far-reaching
coverage of the interconnected networks have facilitated flawless copying, manipulations and
distribution of the digital multimedia (digital video, audio, text, and images). The ever-advancing
storage and retrieval technologies have also smoothed the way for large-scale multimedia
database applications. However, abuses of these facilities and technologies pose pressing threats
to multimedia security management in general, and multimedia copyright protection and content
integrity verification in particular. Although cryptography has a long history of application to
information and multimedia security, the undesirable characteristic of providing no protection to
the media once decrypted has limited the feasibility of its widespread use. For example, an
adversary can obtain the decryption key by purchasing a legal copy of the media but then
redistribute the decrypted copies of the original .In response to these challenges; digital
watermarking techniques have been proposed in the last decade. Digital watermarking is the
procedure whereby secret information (the watermark) is embedded into the host multimedia
content, such that it is: (1) hidden, i.e., not perceptually visible; and (2) recoverable, even after
the content is degraded by different attacks such as filtering, JPEG compression, noise, cropping
etc. The two basic requirements for an effective watermarking scheme, imperceptibility and
robustness, conflict with each other.
The main focus of this thesis is to provide good tradeoff between perceptual quality of
the watermarked image and its robustness against different attacks. For this purpose, we have
discussed two robust digital watermarking techniques in discrete wavelet (DWT) domain. One is
fusion based watermarking, and other is spread spectrum based watermarking. Both the
techniques are image adaptive and employ a contrast sensitivity based human visual system
(HVS) model. The HVS models give us a direct way to determine the maximum strength of
watermark signal that each portion of an image can tolerate without affecting the visual quality
of the image.
In fusion based watermarking technique, grayscale image (logo) is used as watermark. In
watermark embedding process, both the host image and watermark image are transformed into
DWT domain where their coefficients are fused according to a series combination rule that take
into account contrast sensitivity characteristics of the HVS. The method repeatedly merges the
iv
watermark coefficients strongly in more salient components at the various resolution levels of
the host image which provides simultaneous spatial localization and frequency spread of the
watermark to provide robustness against different attacks. Watermark extraction process requires
original image for watermark extraction.
In spread spectrum based watermarking technique, a visually recognizable binary image
is used as watermark. In watermark embedding process, the host image is transformed into DWT
domain. By utilizing contrast sensitivity based HVS model, watermark bits are adaptively
embedded through a pseudo-noise sequence into the middle frequency sub-bands to provide
robustness against different attacks. No original image is required for watermark extraction.
Simulation results of various attacks are also presented to demonstrate the robustness of
both the algorithms. Simulation results verify theoretical observations and demonstrate the
feasibility of the digital watermarking algorithms for use in multimedia standards.





















v
LIST OF FIGURES
1.1 A Digital Watermarking System 3
1.2 Mutual dependencies between the basic requirements .. 7
3.1 The Proposed Fusion-Based Watermark Embedding Method .. 25
3.2 Segmentation of the Host Image Wavelet Coefficients into
wy wx
N N Blocks for
Fusion Watermarking . 26
3.3 Results for Fusion-Based Watermarking Method Without any Attack.. 30
3.4 Results for JPEG Compression .. 31
3.5 Results for Additive White Gaussian Noise Degradation .. 32
3.6 Results for Median Filtering .. 33
3.7 Results for Mean Filtering . 34
3.8 Results for Cropping .. 35
3.9 Results for Image Resizing 36
4.1 Results for Spread Spectrum-Based Watermarking Method Without any Attack 43
4.2 Results for JPEG Compression .. 45
4.3 Results for Additive White Gaussian Noise Degradation .. 46
4.4 Results for Median Filtering .. 47
4.5 Results for Gaussian low pass Filtering . 48
4.6 Results for Cropping .. 49

















vi
LIST OF ACRONYMS
ACF Auto Covariance Function
AWGN Additive White Gaussian Noise
BER Bit Error Rate
CD Compact Disc
EZW Embedded Zero Wavelet Tree
DAB Digital Audio Broadcasting
DCT Discrete Cosine Transform
DFT Discrete Fourier Transform
DSP Digital Signal Processing
DWT Discrete Wavelet Transform
HVS Human Visual System
IDWT Discrete Wavelet Transform
JND Just Noticeable Difference
JPEG Joint Photographic Experts Group
LSB Least Significant Bit
MPEG Moving Picture Experts Group
MSE Mean Square Error
PN Pseudo Noise
PSNR Peak Signal to Noise Ratio
QF Quality Factor
SNR Signal to Noise Ratio





CHAPTER 1








INTRODUCTION

Introduction
Watermarking Requirements
Watermarking Applications
Contribution of the thesis and Chapter organization






1
1.1 INTRODUCTION
In recent years, digital multimedia technology has shown a significant progress. This
technology offers so many new advantages compared to the old analog counterpart. The
advantages during the transmission of data, easy editing any part of the digital content, capability
to copy a digital content without any loss in the quality of the content and many other advantages
in DSP, VLSI and communication applications have made the digital technology superior to the
analog systems. Particularly, the growth of digital multimedia technology has shown itself on
Internet and wireless applications. Yet, the distribution and use of multimedia data is much easier
and faster with the great success of Internet. The great explosion in this technology has also
brought some problems beside its advantages. However, abuses of these facilities and
technologies pose pressing threats to multimedia security management in general, and
multimedia copyright protection and content integrity verification in particular. Although
cryptography has a long history of application to information and multimedia security, the
undesirable characteristic of providing no protection to the media once decrypted has limited the
feasibility of its widespread use. For example, an adversary can obtain the decryption key by
purchasing a legal copy of the media but then redistribute the decrypted copies of the original .In
response to these challenges, digital watermarking schemes have been proposed in the last
decade.
A watermark [1], a secret imperceptible signal, is embedded into the original data in such
a way that it remains present as long as the perceptible quality of the content is at an acceptable
level. The owner of the original data proves his/her ownership by extracting the watermark from
the watermarked content in case of multiple ownership claims. Digital watermark may be
comprised of copyright or authentication codes, or a legend essential for signal interpretation.
The existence of these watermarks with in a multimedia signal goes unnoticed except when
passed through an appropriate detector. Common types of signals to watermark are still images,
audio, and digital video.
As an example of the usefulness of watermarking, let us consider a simple scenario:
Newspaper X publishes a photograph, for which it claims exclusive rights. Newspaper Y, also
claiming to be the exclusive owner, publishes the same photograph after copying it from X.
Without any special protection mechanism, X cannot prove that it is the rightful owner of the
photograph. However, if X watermarks the photograph before publication (that is, X embeds a
2
hidden message that identifies it as its legitimate owner), and is able to detect the watermark later
in the illegally distributed copy, it will be able to supply proof of ownership in a court of law. On
the other hand, to prevent detection of the watermark, Y may try to remove it from the picture by
distorting the picture. That is, Y may attempt to attack the watermark so as to render it
undetectable, without significantly degrading the quality of the image or affecting its commercial
value. Careful design of the watermarking system can prevent this from happening. There have
been many instances of disputes or litigations on the intellectual ownership of multimedia data.
A copyright violations lawsuit that received extensive publicity in the early 2000s, was that
against Napster .Napster was essentially a centralized database which allowed millions of users
to freely distribute music files in a peer-to-peer network. The music files were un-watermarked
and compressed in such a way that the quality of the reproduced music was very close to that of
a Compact Disc (CD recording). However, all copyright information that normally accompanies
the music written on a CD was lost. As a result, it was not an easy task for the music companies
to prove that unauthorized distribution was indeed taking place through Napster. A watermarking
scheme robust to compression would have provided additional ammunition to the music
industry, as the copyright information would have been inseparable from the music itself. Due to
its significance, the watermarking field has grown tremendously over the last years. There are
numerous articles [2, 3, 4, 5, and 6] that explain the basics of watermarking, explore its practical
applications, and evaluate the performance of various schemes under a variety of attacks.
1.2 WATERMARKING SYSTEM
In this thesis, work has been carried out on digital watermarking. Throughout the rest of the
report, watermarking refers to digital watermarking. To avoid the unauthorized distribution of
images or other multimedia property, various solutions has been proposed. Most of them make
unobservable modifications to images that can be detected afterwards. Such image changes are
called watermarks. Watermarking is defined as adding (embedding) a watermark signal to the
host signal. The watermark can be detected or extracted later to make an assertion about the
object. A general scheme for digital watermarking is given in Figure 1.1. The watermark
message can be a logo picture, sometimes a visually recognizable binary picture or it can be
binary bit stream. A watermark is embedded to the host data by using a secret key at the
embedder.

3








The information embedding routine imposes small signal changes, determined by the key and
watermark, to generate the watermarked signal. Only the owner of the data knows the key and it
is not possible to remove the message from the data without the knowledge of the key. Then, the
watermarked image passes through the transmission channel. The transmission channel includes
the possible attacks, such as lossy compression, geometric distortions, any common signal
processing operation and digital-analog and analog to digital conversion, etc. After the
watermarked image passes through these possible operations, the message is tried to be extracted
at the watermark detector. The decoding process can itself performed in two different ways. In
one process the presence of the original unwatermarked data is required and other blind decoding
is possible. The extracted watermark is compared with the original watermark (i.e. the watermark
that was initially embedded) by a comparator function and binary output decision is generated.
The comparator is basically a correlator. Depending on the comparator output it can be
determined if the data is authentic or not. If the comparator output is greater than equal to a
threshold then the data is authentic else it is not authentic.
A watermark is detectable or extractable to be useful. Depending on the way the
watermark is inserted and depending on the nature of the watermarking algorithm, the method
used can involve very distinct approaches. In some watermarking schemes, a watermark can be
extracted in its exact form, a procedure we call watermark extraction. In other cases, we can
detect only whether a specific given watermarking signal is present in an image, a procedure we
call watermark detection. It should be noted that watermark extraction can prove ownership
whereas watermark detection can only verify ownership.
Degraded
Watermarked
Image
Watermarked
Image
Watermark
Embedder
Watermark
Detector
Insecure
Channel
(Attacks)
Host
Image
Detected
Watermark
Message
Host
Image
Watermark
Message
Key
Fig. 1.1 A Digital Watermarking System
4

1.3 WATERMARKING REQUIREMENTS
Watermark by itself is not sufficient to prevent abuses unless a proper protection protocol is
established. The exact properties that a watermarking algorithm must satisfy cannot be defined
exactly without considering the particular application scenario; the algorithm has to be used in.
For example, in the video indexing application, evaluating the robustness of a watermarking
scheme to any signal processing is meaningless, since there is no case that the video passes
through some signal processing operation. In the covert communication application, it is better to
use a watermarking scheme that does not need the original data during the watermark detection
process, if real TV broadcasting is used as the communication channel, while most of the
watermarking schemes in other applications need the original data during the detection process.
If the application is the copyright protection, the owner of the original data may wait for several
days to insert/detect watermark, if the data is valuable for the owner. On the other hand, in a
broadcast monitoring application, the speed of the watermark detection algorithm should be as
fast as the speed of real time broadcasting. As a result, each watermarking application has its
own requirements and the efficiency of the watermarking scheme should be evaluated according
to these requirements.
Each watermarking application has its own specific requirement and the design is
complicated by the conflicting interdependence of the different requirements. It makes it difficult
to study all aspects simultaneously but it appears also hard to successfully isolate the different
constraints. The main requirements which should be fulfilled by a watermarking scheme are
imperceptibility, robustness, capacity
1.3.1 Imperceptibility
Watermarking algorithm must embed the watermark such that this does not introduce any
perceptible artifacts into the host data and not degrade the perceived quality of the underlying
host data. A watermark-embedding procedure is truly imperceptible if humans cannot distinguish
the original data from the data with the inserted watermark [2]. Even the smallest modification in
the host data may become apparent, however, when the original data is compared directly with
the watermarked data. Since users of watermarked data normally do not have access to the
original data, they cannot perform this comparison. Therefore, it may be sufficient that the
5
modifications in the watermarked data go unnoticed as long as the data are not compared with
the original data [6].
1.3.2 Robustness
Robustness refers to the ability to detect the watermark, even if the quality of the host data is
degraded, intentionally (malicious) or unintentionally (non-malicious). In general, there should
be no way in which the watermark can be removed or altered without sufficient degradation of
the perceptual quality of the host data so as to render it unusable.
The Exact level of robustness the hidden data must posses cannot be specified without
considering a particular application. Qualitative robustness level encompassing most of the
situations encountered in practice have been discussed below.
Secure Watermarking:
In this case, mainly dealing with copyright protection, ownership verification or any other
security-oriented application, the watermark must survive both no-malicious as well as malicious
manipulations. In secure watermarking, the loss of the hidden data should be obtainable only at
the expense of a significant degradation of the quality of the host signal. When considering
malicious manipulation it has to be assumed that attackers known the watermarking algorithm
and thereby they can conceive ad-hoc watermark removal strategies. The security must lie on the
choice of key. The watermarking algorithm has truly secure if knowing the exact algorithms for
embedding and extracting the watermark does not help unauthorized party to detect the presence
of the watermark. As to non- malicious manipulations, they include a huge variety of digital and
analog processing tools, including lossy compression, linear and non-linear filtering, cropping
editing, scaling, D/A and A/D conversions, analog duplications, noise addition, and many others
that apply only to particular type. Thus in the image case, we must considering zooming and
shrinking, rotation, contrast, enhancement histogram manipulation, row/ column removal or
exchange, in the case of video we must taken into account frame removal, frame exchange,
temporal filtering, temporal re-sampling, finally robustness of an audio watermark, may imply
robustness against echo addition, multi-rate processing, and pitch scaling. It is though important
to point out that even the most secure system does not need to perfect the contrary, it is only
needed that a high enough degree of security is reached. In other words, watermark breaking
6
does not need to be impossible (which probably will never be the case), but only difficult
enough.

Robust watermarking:
In this case it is required that the watermark be resistant only against non-malicious
manipulations. Robust watermarking is less demanding than secure watermarking. Application
fields in robust watermarking include all the situations in which it is unlikely that someone
purposely manipulates the host data with the intention to remove the watermark. The application
scenario is such that the normal use of data comprise of several kinds of manipulations, which
must not damage the hidden data. Even in copyright protection applications, the adoption of
robust watermarking instead of secure watermarking may be allowed due to the use of a
copyright protection protocol in which all the involved actors are not interested in removing the
watermark.
Semi-fragile watermarking:
Watermark is semi-fragile if it survives a limited well specified, set of manipulations, leaving the
quality of the host document virtually intact. In some applications robustness is not a major
requirement, mainly because the host signal is not intended to undergo any manipulations, but a
very limited number of minor modifications such as moderate lossy compressions, or quality
enhancement. This is the case of data labeling for improved actual retrieval, in which the hidden
data is only needed to retrieve the host data from archive, and thereby it can be discarded once
the data has been correctly assessed. Usually data is archived in compressed format, and that the
watermark is embedded prior to compression. In this case the watermark needs to be robust
against lossy coding.
Fragile watermarking:
A watermark is said to be fragile if the information hidden with in the host data is lost or
irremediably altered as soon as any modification is applied to the host signal. Such a loss of
information may be global, i.e. no part of watermarking can be recovered, or local i.e. only part
of the watermark is damaged. The main application of fragile watermarking is data
authentication, where watermark loss or alternation is taken as evidence that the data has been
7
tampered with. The recovery of the information content within the data demonstrates authentic
un-tampered data.
Robustness against signal distortion is better achieved if the watermark is placed in
perceptually significant part of the signal. This is particularly evident in the case of lossy
compression algorithm, which operates by discarding perceptually insignificant data. Watermark
hidden within perceptually insignificant data are likely not to survive compression. Achieving
watermark robustness, and, to a major extent, watermark security is one of the main challenges
watermarking researches are facing with. Nevertheless its importance has sometimes been over
estimated at the expense of other very important issues as watermark capacity and protocol level
analysis.
1.3.3 Capacity
The capacity requirement of the watermarking scheme refers to be able to verify and distinguish
between different watermarks with a low probability of error as the number of differently
watermarked versions of an image increases [7].
The requirements listed above are all related to each other. The mutual dependencies
between the basic requirements are shown in Fig. 1.2. For instance, a very robust watermark can










be obtained by making many large modifications to the host data for each bit of the watermark.
Large modifications in the host data will be noticeable, however, and many modifications per
watermark bit will limit the maximum amount of watermark bits that can be stored in a data
object. The robustness of the watermarking method increases, the capacity also increases where
Fig. 1.2 Mutual dependencies between the basic requirements
Quality
Capacity
Robustness
8
the imperceptibility decreases. The security of a watermark influences the robustness
enormously. If a watermark is not secure, it cannot be a very robust. Hence, a tradeoff should be
considered between the different requirements so that an optimal watermark for each application
can be developed.
1.4 WATERMARKING APPLICATIONS
Although the main motivation behind the digital watermarking is the copyright protection, its
applications are not that restricted. There is a wide application area of digital watermarking,
including broadcast monitoring, fingerprinting, authentication and covet communication [5, 8].
For secure applications a watermark is used for following purposes:
1. Copyright Protection: For the protection of intellectual property, the data owner can
embed a watermark representing copyright information in his data. This watermark can
prove his ownership in court when someone has infringed on his copyrights.
2. Fingerprinting: To trace the source of illegal copies, the owner can use a fingerprinting
technique. In this case, the owner can embed different watermarks in the copies of the
data that are supplied to different customers. Fingerprinting can be compared to
embedding a serial number that is related to the customers identity in the data. It enables
the intellectual property owner to identify customers who have broken their license
agreement by supplying the data to third parties.
3. Broadcast Monitoring: By embedding watermarks in commercial advertisements, an
automated monitoring system can verify whether advertisements are broadcasted as
contracted. Not only commercials but also valuable TV products can be protected by
broadcast monitoring. The same process can also be used for video and sound clips.
Musicians and actors may request to ensure that they receive accurate royalties for
broadcasts of their performances.
4. Data Authentication: The authentication is the detection of whether the content of the
digital content has changed. As a solution, a fragile watermark embedded to the digital
content indicates whether the data has been altered. If any tampering has occurred in the
content, the same change will also occur on the watermark. It can also provide
information about the part of the content that has been altered
9
5. Copy Protection: The information stored in a watermark can directly control digital
recording devices for copy protection purposes. In this case, the watermark represents a
copy-prohibit bit and watermark detectors in the recorder determine whether the data
offered to the recorder may be stored or not.
6. Covert Communication: The watermark, secret message, can be embedded imperceptibly
to the digital image or video to communicate information from the sender to the intended
receiver while maintaining low probability of intercept by other unintended receivers.
For non-secure applications a watermark is used for following purposes:
1. Indexing: Indexing of video mail, where comments can be embedded in the video
content; indexing of movies and news items, where markers and comments can be
inserted that can be used by search engines.
2. Medical Safety: Embedding the date and the patients name in medical images could be a
useful safety measure.
3. Data Hiding: Watermarking techniques can be used for the transmission of secrete private
messages. Since various governments restrict the use of encryption services, people may
hide their messages in other data.
Although not yet widely recognized as such, bandwidth-conserving hybrid transmission is yet
another information embedding application, offering the opportunity to re-use and share existing
spectrum to either backwards-compatibility increase the capacity of an existing communication
network, i.e., a legacy network, or allow a new network to be backwards-compatibility
overlaid on top of the legacy network. In this case the host signal and embedded signal are two
different signals that are multiplexed, i.e., transmitted simultaneously over the same channel in
the same bandwidth, the host signal being the signal corresponding to the legacy network. Unlike
in conventional multiplexing scenarios, however, the backwards compatibility requirement
imposes a distortion constraint between the host and composite signals.
So-called hybrid in-band on-channel digital audio broadcasting (DAB) is an example of
such a multimedia application where one may employ information embedding methods to
backwards-compatibility upgrade the existing commercial broadcast radio system. In this
application one would like to simultaneously transmit a digital signal with existing analog (AM
and/or FM) Commercial Broadcast radio without interfering with conventional analog reception.
Thus, the analog signal is host signal, and the digital signal is the watermark. Since embedding
10
does not degrade the host signal too much, conventional analog receivers can demodulate the
analog host signal. This embedded signal may be all or part of a digital audio signal, an
enhancement signal used to refine the analog signal, or supplemental information such as station
identification.
1.5 CONTRIBUTION OF THE THESIS AND CHAPTER ORGANIZATION
This thesis addresses the issues regarding Digital watermarking and its applications. The work
included in this thesis aims to provide a good trade-off between perceptual quality of the
watermarked image and its robustness to different attacks, by developing adaptive watermarking
algorithms using wavelet transform and human visual system (HVS) model. Two watermarking
algorithms have been discussed in this work. The first one is multiresolution fusion based
watermarking. The second one is spread spectrum based watermarking technique. The rest of the
thesis is organized as follows:
The different watermarking attacks, and performance measurements which evaluates the
watermarking algorithms is presented in Chapter 2. The literature review on digital
watermarking is also summarized in this chapter.
Fusion based watermarking technique has been discussed in Chapter 3.The technique
requires host signal for watermark extraction and employs image fusion principle to embed both
small grayscale and binary watermarks. Simulation and analysis demonstrates the improved
performance of the technique to a wider variety of attacks such as JPEG compression, filtering,
additive noise, and cropping.
Spread spectrum based watermarking technique has been discussed in Chapter 4. The
technique is a blind technique and watermark bits are embedded through a pseudo noise
sequence. Simulation and analysis demonstrates the robustness of the technique to variety of
attacks.
Finally Chapter 5 presents the concluding remark, with scope for further research work.






CHAPTER 2











WATERMARKING ATTACKS AND
PERFORMANCE MEASUREMENTS

Introduction
Classificationofattacks
Performancemeasuresofwatermarkingalgorithms
Literaturereview
Chaptersummary






11
2.1 INTRODUCTION
To win each campaign, a general needs to know about both his opponents as well as his own
troops. Attacks aim at weakening the watermarking algorithm. The purpose of any watermark-
embedding algorithm is to provide some degree of security and the purpose of any attack is to
negate that purpose. Hence the compilation of a report on watermarking is incomplete without a
mention of watermarking attacks. Study of watermarking algorithm enable to:
Identify weakness of the watermarking algorithm
Propose improvement of the watermarking algorithm
Study effects of current technology on watermark
In watermarking terminology, an attack is any processing that may impair detection of the
watermark or communication of the information conveyed by the watermark. The processed
watermarked data is then called attacked data.
Watermarking is treated as a communication problem, in which the owner attempts to
communicate over a hostile channel, where the non-intentional and the intentional attacks from
the channel. The owner tries to communicate as much watermark information as possible while
maintaining a sufficient high data quality, contrary, and an attacker tries to impair watermark
communication while impairing the data quality as little as possible. Therefore, digital
watermarking scenarios can be considered as a game between the owner and attacker.
Continuing with the analogy of watermarking as a communication system, some researchers
have chosen to work on modeling and resisting attacks on the watermark. They work on the
philosophy that the more specific the information known about the possible attacks, the better we
can design systems to resist it.
2.2 CLASSIFICATION OF ATTACKS
Attacks can be broadly classified as non-malicious (unintentional) such as compression of a
legally obtained, watermarked image or video files and malicious such as an attempt by a
multimedia pirate to destroy the embedded information and prevent tracing of illegal copies of
watermarked digital video. Watermarking systems utilized in copy protection or data
authentication schemes are especially susceptible to malicious attacks. Non-malicious attacks
usually come from common signal processing operations done by legitimate users of the
watermarked materials.
12
2.2.1 Malicious attacks
An attack is said to be malicious if its main goal is to remove or make the watermark
unrecoverable. Malicious attacks can be further classified into two different classes.
Blind: A malicious attack is said to be blind if it tries to remove or make the watermark
unrecoverable without exploiting knowledge of the particular algorithm that was used for
watermarking the asset. For example, copy attack that estimates the watermark signal with aim
of adding it to another asset.
Informed: A malicious attack is said to be informed if it attempts to remove or make the
watermark unrecoverable by exploiting knowledge of the particular algorithm that was used for
watermarking the asset. Such an attack first extracts some secrete information about the
algorithm from publicly available data and then based on this information nullifies the
effectiveness of the watermarking system.
Examples of malicious attacks:
Printing and Rescanning
Watermarking of watermarked image (re-watermarking)
Collusion: A number of authorized recipients of the image should not be able to come
together (collude) and like the differently watermarked copies to generate an un-
watermarked copy of the image (by averaging all the watermarked images).
Forgery: A number of authorized recipients of the image should not be able to collude to
form a copy of watermarked image with the valid embedded watermark of a person not in
the group with an intention of framing a 3rd party.
IBM attack [9]: It should not be possible to produce a fake original that also performs as
well as the original and also results in the extraction of the watermark as claimed by the
holder of the fake original.
2.2.2 Non-Malicious attacks
An attack is said to be non-malicious if it results from the normal operations that watermarked
data or any data for that matter has to undergoes, like storage, transmission or fruition. The
nature and strength of these attacks are strongly dependent on the application for which the
watermarking system is devised.

13
Examples of non-malicious attacks:
Lossy Compression: This is generally an unintentional attack which appears very often in
multimedia applications. Practically all the audio, video and images that are currently
being distributed via Internet have been compressed. If the watermark is required to resist
different levels of compression, it is usually advisable to perform the watermark insertion
task in the same domain where the compression takes place. Many compression schemes
like JPEG and MPEG can potentially degrade the datas quality through irretrievable loss
of data.
Geometric Distortions: Geometric distortions are specific to images videos and include
such operations as rotation, translation, scaling and cropping.
Common Signal Processing Operations: Common signal processing operation includes
such operations such as linear filtering such as high pass and low pass filtering, non linear
filtering such as median filtering, D/A Conversion, A/D conversion, re-sampling, re-
quantization, dithering distortion, addition of a constant offset to the pixel values,
addition of Gaussian and Non Gaussian noise, local exchange of pixels.
The existing attacks can be categorized into four classes of attacks [10]: removal attacks,
geometric attacks, cryptographic attacks, and protocol attacks.
2.2.3 Removal attacks
Removal attacks aim at the complete removal of the watermark information from the
watermarked data without cracking the security of the watermarking algorithm, e.g., without the
key used for watermark embedding. That is, no processing, even prohibitively complex, can
recover the watermark information from the attacked data. This category includes denoising,
quantization (e.g., for compression), re-modulation, and collusion attacks. Not all of these
methods always come close to their goal of complete watermark removal, but they may
nevertheless damage the watermark information significantly. Sophisticated removal attacks try
to optimize operations like de-noising or quantization to impair the embedded watermark as
much as possible while keeping the quality of the attacked document high enough. Usually,
statistical models for the watermark and the original data are exploited within the optimization
process. Collusion attacks are applicable when many copies of a given data set, each signed with
a key or different watermark, can be obtained by an attacker or a group of attackers. In such a
14
case, a successful attack can be achieved by averaging all copies or taking only small parts from
each different copy. Recent results show that a small number of different copies, e.g., about 10,
in the hand of one attacker can lead to successful watermark removal.
2.2.4 Geometric attacks
In contrast to removal attacks, geometric attacks do not actually remove the embedded
watermark itself, but intend to distort the watermark detector synchronization with the embedded
information. The detector could recover the embedded watermark information when perfect
synchronization is regained. However, the complexity of the required synchronization process
might be too great to be practical. For image watermarking, the most known benchmarking tools,
Unzign and Stirmark, integrate a variety of geometric attacks. Unzign introduces local pixel
jittering and is very efficient in attacking spatial domain watermarking schemes. Stirmark
introduces both global and local geometric distortions. We give a few more details about these
attacks later in this paper. However, most recent watermarking methods survive these attacks due
to the use of special synchronization techniques. Robustness to global geometric distortions often
relies on the use of either a transform invariant domain (Fourier-Melline) or an additional
template or of specially designed periodic watermarks whose auto-covariance function (ACF)
allows estimation of the geometric distortions. However, as will be discussed below, the attacker
can design dedicated attacks exploiting knowledge of the synchronization scheme. Robustness to
global affine transformations is more or less a solved issue. However, resistance to the local
random alterations integrated in Stirmark still remains an open problem for most commercial
watermarking tools. The so-called random bending attack in Stirmark exploits the fact that the
human visual system is not sensitive against local shifts and affine modifications. Therefore,
pixels are locally shifted, scaled, and rotated without significant visual distortion. However, it is
worth noting that some recent methods are able to resist against this attack.
2.2.5 Cryptographic attacks
Cryptographic attacks aim at cracking the security methods in watermarking schemes and thus
finding a way to remove the embedded watermark information or to embed misleading
watermarks. One such technique is the brute-force search for the embedded secret information.
Another attack in this category is the so-called Oracle attack, which can be used to create a non-
watermarked signal when a watermark detector device is available. Practically, application of
these attacks is restricted due to their high computational complexity.
15
2.2.6 Protocol attacks
Protocol attacks aim at attacking the entire concept of the watermarking application. One type of
protocol attack is based on the concept of invertible watermarks [9]. The idea behind inversion is
that the attacker subtracts his own watermark from the watermarked data and claims to be the
owner of the watermarked data. This can create ambiguity with respect to the true ownership of
the data. It has been shown that for copyright protection applications, watermarks need to be
non-invertible. The requirement of non-invertibility of the watermarking technology implies that
it should not be possible to extract a watermark from a non-watermarked document. A solution
to this problem might be to make watermarks signal-dependent by using one-way functions.
Another protocol attack is the copy attack. In this case, the goal is not to destroy the watermark
or impair its detection, but to estimate a watermark from watermarked data and copy it to some
other data, called target data. The estimated watermark is adapted to the local features of the
target data to satisfy its imperceptibility. The copy attack is applicable when a valid watermark in
the target data can be produced with neither algorithmic knowledge of the watermarking
technology nor the knowledge of the watermarking key. Again, signal-dependent watermarks
might be resistant against the copy attack.
2.3 PERFORMANCE MEASURES OF WATERMARKING ALGORITHMS
The success of watermarking algorithm is evaluated based on a series of measures [11]. Because
of the psychological nature of the problem not all criteria are quantitative in nature. Although
only some factors are appropriate for a given application, we present all the most popular metrics
below to highlight the character of good watermarking scheme. Without loss of generality, we
assume the host and watermarked signals are images.
1. Perceptual Quality: Perceptual quality refers to the imperceptibility of embedded
watermark data within the host signal. In most applications, it is important that the
watermark is undetectable to a listener or viewer. This ensures that the quality of the host
signal is not perceivably distorted; the peak signal-to-noise ratio (PSNR) of the
watermarked signal versus the host signal was used as a quality measure. The PSNR is
defined as :
16

= =
=
|
|

\
|
=
M
j
N
k
w
n m X n m X
MN
MSE
MSE
PSNR
1 1
2
2
10
)) , ( ) , ( (
1
255
log 10
2.1
in units of dB, where X is host signal, wis the watermark
w
X is the watermarked signal
MN , is the total number of pixels in X or
w
X .
2. Correlation Coefficients: To measure the similarity between embedded and extracted
watermarks, the following normalized correlation coefficients is defined as:
|

\
|
|

\
|
=


m n m n
m n
n m w n m w
n m w n m w
r
) . ( ) , (
) , ( ) , (
2 2
2.2
where w and w are the embedded and extracted watermarks, respectively.
3. Bit Rate: Bit rate refers to the amount of watermark data that may be reliably embedded
within a host signal per unit of time or space, such as bits per second or bits per pixel. A
higher bit rate may be desirable in some applications in order to embed more copyright
information. In this study, reliability was measured as the bit error rate (BER) of
extracted watermark data. For embedded and extracted watermark sequences of length B
bits, the BER (in percent) is given by the expression as:

=
1
0
) ( ) ( 0
) ( ) ( 1
100
B
n
n w n w
n w n w
B
BER 2.3
4. Computational Complexity: Computational complexity refers to the processing
required to embed watermark data into a host signal, and / or to extract the data from the
signal. Algorithm complexity is important to know, for it may influence the choice of
implementation structure or DSP architecture. Although there are many ways to measure
complexity, such as complexity analysis (or Big-0 analysis), for practical applications
more quantitative values are required.



17
2.4 LITERATURE REVIEW
Digital watermarking is a prominent field of research and many researchers have suggested a
large number of algorithms and compared. The main thrust on all such algorithms is to hide
secrete information (watermark) in host signal in such a way that it provides good tradeoff
between imperceptibility and robustness against different attacks. This section presents several
types of digital watermarking techniques found in the academic literature. We do not give an
exhaustive review of the area, but provide an overview of established approaches. Existing
digital watermarking techniques are broadly classified into two categories depending on the
domain of watermark insertion: spatial domain and frequency domain techniques.
The earlier watermarking techniques are almost spatial based approach. In spatial domain
the watermark is embedded into the host image by directly modifying the pixel values, i.e.
simplest example is to embed the watermark in the least significant bits (LSBs) of image pixels
[1]. Spatial domain watermarking is easy to implement and requires no original image for
watermark detection. However, it often fails under signal processing attacks such as filtering and
compression and having relative low-bit capacity. A simple image cropping operation may
eliminate the watermark. Besides, the fidelity of the original image data can be severely
degraded since the watermark is directly applied on the pixel values.
In contrast to the spatial-domain-based watermarking, frequency-domain based
techniques can embed more bits of watermark and are more robust to attack; thus, they are more
attractive than the spatial-domain-based methods, because the watermark information can be
spread out to the entire image. As to the frequency transform, there are DFT (Discrete Fourier
Transform), DCT (Discrete Cosine Transform), and DWT (Discrete Wavelet Transform).
J.J.K.ORunaidh et al. [12] uses phase of the discrete Fourier transform to embed the watermark.
They used the fact that phase is more important than the amplitude of the DFT values for the
intelligibility of an image. Watermarking technique proposed by J.J.K.ORunaidh et al. [13] use
DFT amplitude modulation because of its translation or shift invariant property. Because cyclic
translation of the image in the spatial domain does not affect the DFT amplitude, the watermark
embedded in this domain will be translation invariant. However, embedding watermark in host
image by DFT is suffering from the JPEG attacks. The watermarking technique using DCT and
DWT provides extra robustness to different attacks.
18
I.J. Cox et el. [14] proposed a watermarking technique by taking DCT of entire image. The
method involves adding watermark to the N lowest frequency non-dc DCT coefficients of the
host image where N is the length of the watermark sequence of zero mean and unit variance by
using the following equation:
2.4
where ) , (
2 1
w w F
DCT
and ) , (
2 1
w w F
DCT
w
are the DCT coefficients of the host image and
watermarked image respectively, a is the scaling parameter, and w(i) is the i
th
watermark
element. This algorithm is one of the earliest attempts at providing image adaptability in the
watermark embedding scheme. This is due to the fact that the watermark strength depends on the
intensity of the DCT coefficients of the original image. In this way, the watermark signal can be
quite strong in the DCT coefficients with large intensity values and is attenuated in the areas with
small DCT coefficients. F.M. Boland et el. [15] proposed a method which also modulates the
coefficients but uses a one-dimensional bipolar binary sequence. The marking procedure consists
of sorting the DCT coefficients of the image according to their absolute magnitude. The
watermark is then added to the N largest AC coefficients. Inserting the watermark into the
perceptually significant components and adapting the watermark strength by the strength of the
DCT component provides a watermark that is quite robust and transparent. However, because the
DCT is obtained on the entire image rather than the usual block-based approach commonly
found in image and video compression schemes, the transform does not allow for local spatial
control of the watermark insertion process. In other words, the addition of a watermark value to
one DCT coefficient affects the entire image. The method in M. Barni et el. [16] is a slight
modification of previous work [14], where the authors allow the user to determine the scaling
factor and coefficients to be marked. The user-defined scaling factor and watermark length will
greatly influence the effectiveness of this scheme both in terms of transparency and robustness.
In all [14, 15, 16], original image is required for watermark extraction.
In [17], S. Burgett et el. uses block based DCT approach to embed the watermark. The
image is segmented into 8 8 non-overlapping blocks and the DCT of each block is obtained
similar to JPEG. A random subset of the blocks is chosen and a triplet of midrange frequency
coefficients is slightly altered to encode a binary sequence. This seems to be a reasonable
approach for adding some sort of perceptual criterion. Watermarks inserted into the high
frequencies are most vulnerable to attack, whereas the low-frequency components are
perceptually significant and very sensitive to alterations; such alterations may make the
)) ( 1 )( , ( ) , (
2 1 2 1
i aw w w F w w F
DCT DCT
w
+ =
19
watermark visible. Bors and Pitas [18] suggest a method that modifies DCT coefficients
satisfying a block site selection constraint. The image is first divided into blocks of size 8 8 .
Certain blocks are then selected according to a Gaussian network classifier decision. The middle
range frequency DCT coefficients are then modified, using either a linear DCT constraint or a
circular DCT detection region, to convey the watermark information. In [17, 18], original image
is not required for watermark extraction. This technique provides reasonable results on average,
although a more image-dependent scheme could provide better quality and robustness. Image
adaptive watermarking scheme using HVS model improves the performance of the watermarking
techniques.
Swanson et al. [19] suggest a DCT domain watermarking technique, based on frequency
masking of DCT blocks. The input image is split up into square blocks for which the DCT is
computed. For each DCT block, a frequency mask is computed based on the knowledge that a
masking grating raises the visual threshold for signal gratings around the masking frequency.
The resulting perceptual mask is scaled and multiplied by the DCT of a maximal length PN
sequence. This watermark is then added to the corresponding DCT block followed by spatial
masking to verify that the watermark is invisible and to control the scaling factor. Watermark
detection requires the original image as well as the original watermark and is accomplished by
hypothesis testing. The scheme is robust against JPEG compression, noise, and cropping.
Tao and Dickinson [20] propose an adaptive block based DCT domain watermarking
technique based on a regional perceptual classifier with assigned sensitivity indexes. The
watermark is embedded in N AC DCT coefficients. The coefficients are selected as to have the
smallest quantization step sizes according to the default JPEG compression table. Various
approaches exist to determine the noise sensitivity by efficiently exploiting the masking effects
of the HVS. The authors propose a regional classification algorithm which classifies the block in
one of six perceptual classes. The classification algorithm exploits luminance masking, edge
masking, and texture masking effects of the HVS. Namely the perceptual block classes from one
to six are defined as: edge; uniform; low sensitivity; moderately busy; busy; and very busy, in
descending order of noise sensitivity. Each perceptual class has a noise-sensitivity index
assigned to it. Watermark recovery requires the original image as well as the watermark and is
based on hypothesis testing. The author report shows that the method is robust to JPEG
compression and additive noise.
20
C. Podilchuk, and W. Zeng [21] propose a watermarking technique for digital images that
is based on utilizing visual models, which have been developed in the context of image
compression. The visual model gives a direct way to determine the maximum amount of
watermark signal that each portion of an image can tolerate without affecting the visual quality
of the image. The watermark encoding scheme consists of a frequency decomposition based on a
8 8 framework followed by just noticeable difference (JND) calculation and watermark
insertion. The watermark scheme is robust to different attacks such as JPEG compression,
additive noise, scaling etc.
J. Wu, and J. Xie [22] propose an adaptive watermarking technique in DCT domain using
HVS model and fuzzy c-means technique (FCM). In this method FCM technique is used to
classify non-overlapping 8 8 original blocks into categories: one is suitable for watermarking
with high imperceptibility and robustness and the other is unsuitable. Watermark is inserted in
DCT mid-frequency coefficients of selected blocks. W. Zhang et el. [23] propose an adaptive
digital watermarking approach. In this method FCM technique is used to determine the
watermark strength of each image pixel, and then watermark is inserted adaptively to the N
largest magnitude non-dc DCT coefficients of the host image. The both the method performs
better against additive noise, compression and cropping etc.
Yifei Pu. et el. [24] proposes a public adaptive watermark algorithm for color images
based on principal components analysis of generalized Hebb. The algorithm is based on principal
component analysis of generalized Hebb adaptive algorithm in Artificial Neural Network and to
do adaptive quantitative coding for principal component coefficients according to the proportion
of marginal or textural information of the watermark image. In addition, it adaptively adjusts the
embedding depth according to the images features to ensure the invisibility of the watermark. By
way of disporting and stochastic embedding into color image watermark, it increases the
embedding robusticity of watermark.
Although embedding watermark in host image by DCT is more robust than that of by
DFT, the DWT has a number of advantages over the DCT, because the DWT provides both
space and frequency localization, and different resolution levels. Thus, DWT based
watermarking algorithm can effectively utilize the characteristics of HVS (Human Visual
System) to attain good trade-off between robustness and imperceptibility. So, DWT based
watermarking algorithms have gained more interest among the watermark researchers.
21
X.-G. Xia et el. [25] proposes a multiresolution watermark for digital images. The
technique is implemented in DWT domain and watermark is inserted in the same way as
described in method [15]. C. Podilchuk, and W. Zeng [26] propose image adaptive watermarking
using visual models. The method is implemented using DWT and a HVS model and watermark
is embedded adaptively by calculating just noticeable difference (JND) for each block regions.
M.-S. Hseih [27] proposes a hiding digital watermarks using multiresolution wavelet transform.
In this method original image is decomposed into wavelet coefficients. The method embeds a
visually recognizable binary or gray image by modifying the mid frequency part of the image.
Watermarking methods is based on the qualified significant wavelet tree which comes from the
concept of embedded zero wavelet tree (EZW). The above methods are robust to a variety of
signal distortions and requires original image for watermark extraction. In methods [25, 26, 27]
the watermark embedded linearly to the original image. Deepa Kundur, and D. Hatzinakos [28]
propose a digital watermarking using multiresolution wavelet decomposition. In this method
watermark is embedded non-linearly in the original image by using scalar quantization, and
image fusion principle concept. Original image is not required for watermark extraction. In this
thesis we propose a fusion based image adaptive watermarking method using wavelet transform
and a HVS model based on contrast sensitivity.
In [29] Mauro Barni et el. have proposed a scheme where in contrast to conventional
methods operating in the wavelet domain, masking is accomplished pixel by pixel by taking into
account the texture and the luminance content of all the image sub-bands. The watermark
consists of a pseudorandom sequence which is adaptively added to the largest detain bands. As
usual, the watermark is detected by computing the correlation between the watermarked
coefficients and watermarking code, anyway detection threshold is chosen in such way that the
knowledge of watermark energy used in the embedding phase is not needed, thus permitting to
adapt it to the image at hand.
In [30] Xiangui Kang et el. have proposed a blind discrete wavelet transform- discrete
Fourier transform (DWT-DFT) composite image watermarking algorithm that is robust against
both affine transform and JPEG compression. This algorithm improves the robustness via using
new embedding strategies, watermark structure, 2-D interleaving, and synchronization technique.
A spread spectrum based informative watermark with a training sequence are embedded in the
coefficients of the LL sub-band in the DWT domain while a template is embedded in the middle
frequency components in the DFT domain. In watermark extraction, we first detect the template
22
in a possibly corrupted watermarked image to obtain the parameters of affine transform and
convert the image back to its original shape. Then we perform translation registration by using
the training sequence embedded in the DWT domain and finally extract the informative
watermark.
In [31] Jianzhen wu et el. have proposed a blind wavelet based watermarking scheme
using fuzzy clustering theory. The watermarking scheme utilizes the HVS by clustering the local
image features, and thus can embed more robust watermark under a certain visual distance.
Watermark bits are embedded through a PN sequence. In order to improve the robustness, we
embed watermark several times in different position, which are randomly chosen. Similarly, in
this thesis we propose a spread spectrum based blind image adaptive watermarking method using
wavelet transform and a HVS model based on contrast sensitivity.
In [32] Zhang Guannan et el. have proposed an adaptive block-based blind watermarking
algorithm using DWT. By analyzing the characteristic of detail sub-band coefficients of the
image after discrete wavelet transform, we use the mean and variance of the detail sub-bands to
modify the wavelet coefficients adaptively to embed the watermark. This is a blind watermark
algorithm to confirm the copyright without the original image and the watermark is a meaningful
binary image. The author report concludes that the algorithm is robust to common image
processing operations.
2.5 CHAPTER SUMMARY
The various watermarking attacks in the image processing domain were discussed. Parameters
that measure the performance of watermarking algorithms against different attacks were
presented. The existing watermarking algorithms in different domain found in academic
literature are also surveyed.











CHAPTER 3











FUSION-BASED WATERMARKING

Introduction
Algorithm Description
SimulationResults andDiscussion
ChapterSummary






23
3.1 INTRODUCTION
This chapter presents an approach for robust source extraction watermarking algorithm
based on multi-resolution image fusion principle. We address the problem of embedding binary
images, gray images robustly within the host signal. The method transforms both the host image
and watermark into the discrete wavelet domain where their coefficients are fused according to a
series combination rule that take into account contrast sensitivity characteristics of the HVS [36].
The watermark is restricted to be much smaller in dimension than the host signal. No randomly
generated keys are required for security, but the host image is necessary for watermark
extraction. The method repeatedly merges the watermark coefficients at the various resolution
levels of the host signal which provides simultaneous spatial localization and frequency spread
of the watermark to provide robustness against widely varying signal distortions including
cropping and filtering. The watermarking process is adaptive and depends on the local host
image characteristics at each resolution level. Moreover, the watermark is resilient to attack since
it is embedded strongly in more salient components of the image.
We develop our approach to fulfill the following requirements of a successful robust
watermarking scheme:
1. The data hiding technique is adaptive and takes into account the natural masking
characteristics of the host signal to more strongly, and hence, reliably embed the
watermark.
2. The embedded watermark is robust to a reasonable level of signal distortion. Since the
host signal is available for watermark extraction, it is exploited to characterize any
attacks.
3. The algorithm is portable to different applications and can hide different types of
information robustly within a host signal.
Research into human perceptions indicates that the retina of the eye splits an image into
several components which circulates from the eye to the cortex in differently tuned channels
(frequency bands). These channels can only be excited by the component of a signal with similar
characteristics. The processing of signals in different channels is independent. Studies have
shown that each of these channels have a bandwidth of approximately one octave [33]. Similarly,
in a multi-resolution decomposition, the image is separated into bands of approximately equal
24
bandwidth on a logarithmic scale. It is therefore expected that use of the discrete wavelet
transform will allow the independent processing of the resulting components without significant
perceptible interaction with them.
For this reason, wavelet decomposition is attractive for the fusion of images. Image
fusion refers to the processing and synergistic combinations of images from various knowledge
sources and sensors to provide an overall result which contains the most relevant characteristics
of its components. Since the process of image fusion is essentially a sensor-compressed
information problem (i.e., it involves the combining of one or more images into a single fused
result), it follows that wavelets are also useful for such merging.
Some multi-resolution wavelet fusion methods make use of information about the HVS to
determine the perceptually most significant information from each image to retain the composite
[34]. It is then expected that such rules can be used to judiciously select the regions of the host
image in which to embed the watermark.
3.2 ALGORITHM DESCRIPTION
Throughout our discussion, we use ) , ( n m X to denote the host image and ) , ( n m w the watermark.
The watermark, assumed to be a two dimensional array of real elements. The watermark is
visually recognizable binary or gray scale image. The size of the watermark is N N . It is required
that the size of the watermark in relation to the host image be small. We assume, without loss
of generality, that the watermark is smaller than the host by a factor of
M
2 , whereM is an integer
greater or equal to 1.
3.2.1 Watermark Embedding Method
The technique is comprised of the 3 main stages is summarized in Figure 3.1. First, the image
and watermark both are decomposed using the DWT. In the second stage, the watermark is
selectively and repeatedly merged using a model of human contrast sensitivity to determine the
most salient localized host image components. Last, the inverse DWT is applied to form the
watermarked image. The following is the more detailed and analytic description of the
procedure.


25








Stage 1:
The host image and the watermark are transformed into the wavelet domain. We perform the
th
L
level DWT of the host image to produce a sequence of L 3 detail images, corresponding to the
horizontal, vertical, diagonal details at each resolution levels, and a gross approximation image at
the coarsest resolution level. The value of L is equal to 1 + M . We denote the
th
k detail image
component at the
th
l resolution level of the host by ) , (
,
n m X
l k
, where 3 , 2 , 1 = k represents the
frequency orientation corresponding to the horizontal, vertical and diagonal image details,
L l ,...., 1 = the resolution level and ) , ( n m particular pair spatial location index at the resolution l
.The gross approximation is represented by ) , (
, 4
n m X
L
where the subscript 4 is used instead of
k to denote the gross image approximation at resolution L .
Similarly, the first level DWT of the watermark w is performed to produce
wy wx
N N
dimensional detail and approximation sub-images denoted by ) , (
1 ,
n m w
k
where 4 , 3 , 2 , 1 = k .
Stage 2:
The each sub-images of the host are segmented into non-overlapping
wy wx
N N blocks. Figure
3.2 demonstrates the procedure. We denote the segments by ) , (
,
n m X
i
l k
where
) 1 ( 2
2 ,...., 2 , 1
l M
i
+
=
is the total number of blocks at each frequency orientation k and resolutionl .
The salience, S (which is numerical measure of perceptual importance) of each of the
localized blocks is computed using information about the HVS model based on contrast
sensitivity. The value of the salience determines the strength of the watermark to embed in the
particular
wy wx
N N coefficient image block. To define our measure of salience, we first
introduce the notion of contrast sensitivity. Mathematically contrast sensitivity is defined as the

Lth-level
DWT
Multi-resolution
HVS-based
Image Fusion
Watermarked
Image Inverse
Lth-level
DWT
Watermark
Host
Image
1st-level
DWT
Stage 1 Stage 2 Stage 3
Figure 3.1: The Fusion-Based Watermark Embedding Method
26

















reciprocal of the contrast necessary for a given spatial frequency to be perceived. For this paper
we assume the well known model given by Dooley [35]. We extend the model to two
dimensional using the same approach as [34]. The resulting contrast sensitivity for a particular
pair of spatial frequency is given by:
) 1 ( 05 . 5 ) , (
) ( 1 . 0 ) ( 178 . 0
=
+ + v u v u
e e v u C 3.1
Where ) , ( v u C is the contrast sensitivity matrix andu , and v are the spatial frequencies. The
salience of each block is defined as:
2
) , (
, ,
) , ( ) , ( )) , ( (

=
v u
i
l k
i
l k
v u F v u C n m X S 3.2
Where ) , (
,
v u F
i
l k
the normalized discrete Fourier is transform of the image component ) , (
,
n m X
i
l k
;
) , (
,
v u F
i
l k
is normalized such that it has unit energy (i.e. 1 ) , (
2
,
= v u F
i
l k
). The image fusion
method presented relies on the contrast sensitivity of the HVS to determine the importance of the
information.
Watermark wavelet
sub-images of
dimension
wy wx
N N

1 , 1
W
1 , 2
W
1 , 3
W
1 , 4
W
Each rectangular
region contains
wy wx
N N
Coefficients
Embed
1 , 2
W in
Selected blocks
of
l
X
, 2
for all l
Similarly embed
l k
W
,
in
Selected blocks of
l k
X
,
for
4 , 2 , 1 = k and for all l
Segmentation of host signal DWT
Sub-images into
wy wx
N N blocks
Figure 3.2: Segmentation of the Host Image Wavelet Coefficients into
wy wx
N N Blocks for
Fusion Watermarking. The Salience of each block is computed and if it is above a specified
threshold, the corresponding
wy wx
N N watermark wavelet coefficient is embedded. As
suggested by the diagram, the watermark is more widely spread spatially when embedded at a
lower (coarser) resolution
27
The watermark is embedded only in B percent of the most salient detail image blocks at
each resolution level and orientation using the following equation:
) , (
)) , ( ( max
)) , ( (
) , ( ) , (
,
,
,
, ,
,
,
n m w
n m X S
n m X S
n m X n m X
l k
i
l k
i
l k i
l k
i
l k
i w
l k
+ =
3.3
where ) , (
,
,
n m X
i w
l k
are the watermarked DWT coefficients. For the remaining blocks, we set
) , ( ) , (
,
,
,
n m X n m X
i
l k
i w
l k
= 3.4
where
i
l k,
are positive real numbers that determine a tradeoff between the imperceptibility and
robustness against attacks at each of the resolution levels. The value of
i
l k,
is adaptively
changed according to the resolution level. The value of
i
l k,
ranges between 5% and 75% of
the mean value of the detail image blocks. For each resolution levels, the value of
i
l k,
is set
such that lower value for higher resolution level and correspondingly higher value for next lower
resolution levels. The fraction within the square root is a relative measure that gives greater
weight judiciously to the embedded watermark in more salient host image regions.
A similar merging procedure is used to embed the watermark approximation coefficients
) , (
1 , 4
n m w into the host image approximation block ) , (
, 4
n m X
i
L
. The watermark is embedded in all
blocks. The value of
L , 4
is set between 1% and 5% of the mean value of the approximate image
block to ensure imperceptibility.
The larger the magnitude of
l k,
, the more robust and visible the watermark; the ranges of
value suggested provide an appropriate trade-off for most photographic images. Similarly, the
larger the value of B, the greater the number of coefficient blocks in which the watermark is
embedded at each resolution level which also comes at the expense of increased visibility;
simulation results shows that a range of B between 25 and 75 allows for appropriate marking.
Stage 3:
The corresponding
th
L level inverse DWT (IDWT) of the fused image components ) , (
,
n m X
w
l k
is
computed to form the watermarked image.
3.2.2 Watermark Extracting Method
The objective of the extraction process is to reliably obtain an estimate of the original watermark
from a possibly distortion version of the watermarked image
w
X . The reconstruction process
28
requires knowledge of the original host image X . The watermark is extracted from the possibly
corrupted watermarked image using the host image, by applying the inverse procedure at each
resolution level to obtain an estimate of the watermark. The estimates for each resolution level
are averaged to produce an overall estimate of the watermark. The normalized correlation
coefficient r was used to measure the robustness of the extracted watermark against different
attacks.
3.3 SIMULATION RESULTS AND DISCUSSION
For simulations, we take Lena image of size 512 512 as the host image shown in Fig. 3.3(a)
and watermark is visually recognizable gray-scale image of size 32 32 shown in Fig. 3.3(b). To
form the watermark, the DC value is first subtracted from the watermark image and then made
its variance value to 1, before watermark image is used for simulation. We chose 75 = B , ; 5 = L
and value was set to 60, 40, 20, 10, and 5 percent of mean value of detail image blocks for
lower resolution level to higher resolution level respectively, and value was set to 1.6% of
approximate image blocks in our simulation. The PSNR value of watermarked image is 37.5381
as shown in Fig. 3.3(c), and is perceptually identical to the original host and watermark can be
exactly extracted. The resulting watermarked image is corrupted using one of many common
distortions which we discuss in the subsequent section. When the watermark was extracted it was
scaled, so that its minimum pixel value was set to black and its maximum pixel value to white
and correlated with the embedded watermark to measure the robustness and detection capability
of the technique
Robustness against JPEG Lossy Compression
Figure 2(a) shows the effect of compression on the correlation coefficient for different quality
factors. The correlation coefficient remains high for reasonable quality factor values. Severe
visual image degradation in which the features of the face were not distinguishable occurred for
quality factors of 15 and above. The results show that the watermark still remains present and
correlation coefficient is still high about 0.8. Fig. 3.4(b), and 3.4(d) shows the degraded
watermarked image and Fig. 3.4(c) and 3.4(e) shows the corresponding extracted watermark for
quality factor 15 and 5 respectively.


29
Robustness against AWGN Noise
Figure 3.5(a) provides the results for degradation using additive white Gaussian noise. The
proposed method performs well in the presence of additive noise. Severe visual image
degradation occurred at signal to noise ratios of 15 dB and greater. Although the image appeared
overwhelmed by noise, the watermark can be detected with a correlation of about 0.8. Fig.
3.5(b), and 3.5(d) shows the degraded watermarked image and Fig. 3.5(c), and 3.5(e) shows the
corresponding extracted watermark for SNR 15dB and 10dB respectively.
Robustness against Filtering
The results for degradations from median and mean filtering are also presented in Fig. 3.6(a) and
3.7(a) respectively. The watermarked image was filtered with a F F mean (or median) filter.
Highly noticeable image degradation began to occur for 9 > F . The watermark can still be
detected. . Fig. 3.6(b), and 3.6(d) shows the degraded watermarked image and 3.6(c), and 3.6(e)
shows the corresponding extracted watermark for median filtering of order 5 5 and 9 9
respectively. Fig. 3.7(b), and 3.7(d) shows the degraded watermarked image and 3.7(c), and
3.7(e) shows the corresponding extracted watermark for mean filtering of order 5 5 and 9 9
respectively.
Robustness against Cropping
Fig 3.8(a) shows the effect of image cropping on watermark extraction. For watermark
extraction, the portion of the watermarked image cropped out was replaced with the host image.
Even when only 25% of the image area cropped, the correlation value for the proposed technique
is high about 0.9. Fig. 3.8(b), and 3.8(d) shows the degraded watermarked image and Fig. 3.8(c)
and 3.8(e) shows the extracted watermark for 12.5% and 25% image area cropped respectively.
Robustness against Image Resizing
Fig 3.9(a) shows the results of the watermarked images. The images were scaled down in size by
a factor of F using bilinear interpolation and were resized to their original dimension before
watermark extraction. Visible degradation occurs for high value of F due to resolution
adjustment, but the watermark can still be detected with correlation of about 0.8. Fig. 3.9(b), and
3.9(d) shows the degraded watermarked image and Fig. 3.9(c) and 3.9(e) shows the extracted
watermark for image scaling down by a factor of 5 and 7 respectively.
30

























(a) (b)
(c)
Fig.3.3. Results for Fusion-Based Watermarking Method Without any Attack: (a) original
image, (b) watermark image, (c) watermarked image.

31






























(a)
0 10 20 30 40 50 60 70 80
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Quality Factor
c
o
r
r
e
l
a
t
i
o
n

C
o
e
f
f
i
c
i
e
n
t
s
JPEG Compression
(c) (b)
(d)
(e)
Fig.3.4 Results for JPEG Compression, (a)Correlation coefficient vs. Quality factor (QF), (b)
degraded watermarked image for QF=15, (c) extracted watermark for QF=15, (d) degraded
watermarked image for QF =5, (e) extracted watermark for QF=5.
32





















.

























0 5 10 15 20 25 30 35 40 45 50
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
SNR (dB)
c
o
r
r
e
l
a
t
i
o
n

C
o
e
f
f
i
c
i
e
n
t
s
AWGN Noise
(b) (c)
(d) (e)
Fig.3.5 Results for Additive White Gaussian Noise Degradation, (a)Correlation coefficient vs.
SNR, (b) degraded watermarked image for SNR=15dB, (c) extracted watermark for SNR=15dB,
(d) degraded watermarked image for SNR=10dB, (e) extracted watermark for SNR=10dB.

33
















































2 4 6 8 10 12
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Filter Dimension
c
o
r
r
e
l
a
t
i
o
n

C
o
e
f
f
i
c
i
e
n
t
s
Median Filtering
(b) (c)
(d) (e)
Fig.3.6 Results for Median Filtering, (a)Correlation coefficient vs. dimension of filter F, (b)
degraded watermarked image for F=5, (c) extracted watermark for F=5, (d) degraded
watermarked image for F =9, (e) extracted watermark for F=9.

34
















































2 4 6 8 10 12
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Filter Dimension
c
o
r
r
e
l
a
t
i
o
n

C
o
e
f
f
i
c
i
e
n
t
s
Mean Filtering
(a)
(c) (b)
(d)
(e)
Fig.3.7 Results for Mean Filtering, (a)Correlation coefficient vs. dimension of filter F, (b)
degraded watermarked image for F=5, (c) extracted watermark for F=5, (d) degraded
watermarked image for F =9, (e) extracted watermark for F=9.
35
















































Fig.3.8 Results for Cropping, (a) Correlation coefficient vs. percent cropped image area, (b)
degraded watermarked image for 12.5% image area cropped, (c) extracted watermark for
12.5% image area cropped, (d) degraded watermarked image for 25% image area cropped, (e)
extracted watermark for 25% image area cropped.
(a)
0 5 10 15 20 25
0.93
0.94
0.95
0.96
0.97
0.98
0.99
1
cropped image area
c
o
r
r
e
l
a
t
i
o
n

C
o
e
f
f
i
c
i
e
n
t
s
Cropping
(c) (b)
(e) (d)
36
















































1 2 3 4 5 6 7
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Scaling down Factor
c
o
r
r
e
l
a
t
i
o
n

C
o
e
f
f
i
c
i
e
n
t
s
Image resize
(a)
(d) (e)
(b) (c)
Fig.3.9 Results for Image Resizing, (a) Correlation coefficient vs. image scaling down factor,
(b) degraded watermarked image for factor 5, (c) extracted watermark for factor 7, (d)
degraded watermarked image for factor 5, (e) extracted watermark for factor 7.
37
A simulation result shows that, the multiresolution fusion-based watermarking method
performs better against different attacks. The use of the DWT domain inherently makes our
design more resilient to localized spatial and frequency domain distortions including filtering,
resolution reduction and cropping. From experience with other host images, we find that the
method works significantly better for images with highly varying localized characteristics) i.e.,
images with both smooth and busy areas). This is due to the fact that our HVS-based merging
rule adapts the watermark signal strength to the local masking characteristics of the host image.
Thus, a higher energy signal can be imperceptibly embedded with in all regions of the signal. An
advantage of our method is its flexibility in embedding both binary and grayscale logo
watermarks. Experimentally, we found that the embedded watermark undergoes at worst the
same level of perceptible distortion as the watermarked image. This is an inherent advantage to
our fusion-based watermarking scheme since an attacker would have to destroy the watermarked
image to guarantee that the watermark was sufficiently degraded.
3.4 CHAPTER SUMMARY
A multiresolution fusion based watermarking technique employing a model of the HVS was
described in the chapter. The key features of the approach are summarized below:
The watermark is first decomposed using a 1
st
-level DWT so that its detail coefficients
can be repeatedly embedded into the corresponding detail coefficients of the host using a
model of HVS. This process involves merging components of the watermark with similar
characteristics within those of the host image, so that the technique better exploits the
masking properties of the host signal. Hence watermark is embedded with much stronger
energy while remaining imperceptible within the host signal.
The technique embeds the watermark DWT coefficients repeatedly in the discrete
wavelet domain which inherently makes the watermark more resilient to filtering,
cropping, and resolution reduction than techniques using other type of transforms.







CHAPTER 4











SPREAD SPECTRUM-BASED WATERMARKING

Introduction
Algorithm Description
SimulationResults andDiscussion
ChapterSummary







38
4.1 INTRODUCTION
This chapter presents an approach for robust destination extraction watermarking based on
spread spectrum principle. We address the problem of embedding binary data, and visually
recognizable binary images robustly within the host signal. The method transforms the host
signal into the discrete wavelet domain where the watermark bits are embedded through a
pseudo-random white noise sequence in the middle frequency sub-bands to achieve good
tradeoff between robustness and imperceptibility [31]. The watermark bits are adaptively
embedded in the host signal by utilizing the contrast sensitivity characteristics of the human
visual system (HVS). No original image is required for watermark extraction, but only a secret
key is necessary for extraction. Hence watermarking process is more practical. The
watermarking process is adaptive and depends on the local host image characteristics.
Moreover, the watermark is resilient to attack since watermark bit is embedded strongly in more
salient components of the image. The main advantages of the proposed watermarking method are
listed below:
1. The data hiding technique is adaptive and takes into account the natural masking
characteristics of the host signal to more strongly, and hence, reliably embed the
watermark.
2. The data hiding technique is more practical as no original image is required for
watermark extraction.
3. The visually recognizable binary image can be used as watermark to claim ones
ownership.
4.2 ALGORITHM DESCRIPTION
In the proposed watermarking scheme, we use ) , ( n m X to denote the host image and visually
recognizable binary image ) , ( n m w is used as watermark.
4.2.1 Watermark Embedding Method:
The watermark embedding technique is comprised of the 3 main stage discussed below. First, the
image is decomposed using the DWT. In the second stage, the watermark bits are adaptively
embedded through a PN-sequence using a model of human contrast sensitivity. Last, the inverse
DWT is applied to form the watermarked image. The following is the more detailed and analytic
description of the procedure.
39


Stage 1:
The host image is transformed into the wavelet domain. We perform the 1st-level discrete
wavelet decomposition of the original image, and we got 3 detail images, corresponding to the
horizontal, vertical, diagonal details, and 1 gross approximation image. We denote the
th
k detail
image component of the host by ) , (
1 ,
n m X
k
, where 3 , 2 , 1 = k represents the frequency orientation
corresponding to the horizontal, vertical and diagonal image details, and 1 represents the first
resolution level and ) , ( n m particular pair spatial location index. The gross approximation is
represented by ) , (
1 , 4
n m X where the subscript 4 is used instead of k to denote the gross
approximation image. In order to avoid serious image degradation and survive lossy
compression, we will embed the watermark in the middle frequency band that is
) , (
1 , 1
n m X and ) , (
1 , 2
n m X . We split ) , (
1 , 1
n m X and ) , (
1 , 2
n m X sub-band into non-overlapping 8 8
blocks respectively, suppose that the original image is of M M , then
) , (
1 , 1
n m X and ) , (
1 , 2
n m X will be of size
2 2
M M
. After splitting there will be
16 16
M M
blocks
respectively in ) , (
1 , 1
n m X and ) , (
1 , 2
n m X sub-band.
The watermark image is converted into an array of bits. If the watermark is 32 32 , the
number of bits is1024. The number of watermark bits used should be less than total number of
blocks in ) , (
1 , 1
n m X or ) , (
1 , 2
n m X sub-band.
Stage 2:
The salience S (which is a numerical measure of perceptual importance) of each of these
localized segments is computed using information about the contrast sensitivity characteristics of
the HVS. The value of the salience determines the strength of the watermark to embed in the
particular 8 8 coefficient image block. Mathematically, contrast sensitivity is defined as the
reciprocal of the contrast necessary for a given spatial frequency to be perceived.
The salience S of each localized block is determined by the same procedure as described
in chapter 3. Again for convenience the resulting contrast sensitivity for a particular pair of
spatial frequencies is given by:
) 1 ( 05 . 5 ) , (
) ( 1 . 0 ) ( 178 . 0
=
+ + v u v u
e e v u C 4.1
40

where ) , ( v u C is the contrast sensitivity matrix and u and v are the spatial frequencies. The
salience of each block is defined as:
2
) , (
1 , 1 ,
) , ( ) , ( )) , ( (

=
v u
k
i
k
v u F v u C n m X S 4.2
where ) , (
,
v u F
i
l k
the normalized discrete Fourier is transform of the image component ) , (
1 ,
n m X
i
k
;
) , (
1 ,
v u F
i
k
is normalized such that it has unit energy (i.e. 1 ) , (
2
1 ,
= v u F
i
k
). The method presented
relies on the contrast sensitivity of the HVS to determine the importance of the information
In order to keep secret of watermark embedding position, we generate pseudo random number to
be used as the allocation of the watermarking position of the blocks in ) , (
1 , 1
n m X and ) , (
1 , 2
n m X
sub-band. In generating the pseudo random number, a 'key' is used as a seed number. To fit the
random number to the number of blocks in ) , (
1 , 1
n m X and ) , (
1 , 2
n m X , it is scaled to the block
numbers in ) , (
1 , 1
n m X and ) , (
1 , 2
n m X sub-band. Watermark is embedded in chosen blocks in
) , (
1 , 1
n m X and ) , (
1 , 2
n m X only. We use another different key to generate an 8 x 8 random
sequence having distribution of ) 1 , 0 ( N to embed a watermark bit in each chosen block. The same
watermark bit is embedded in the chosen blocks, which have the same location in ) , (
1 , 1
n m X and
) , (
1 , 2
n m X sub-band. Watermark bit embedding procedure can be represented as follows:
)) , ( ( max
)) , ( (
,
,
1 ,
n m X S
n m X S
i
l k
i
l k i
k
= 4.3
If watermark bit=1
) , ( _ ) , ( ) , (
, , ,
,
,
n m one PN n m X n m X
ci
l k
ci
l k
ci
l k
ci w
l k
+ = 4.4
else
) , ( _ ) , ( ) , (
, , ,
,
,
n m one PN n m X n m X
ci
l k
ci
l k
ci
l k
ci w
l k
= 4.5
Where 8 , 1 n m , and 2 , 1 = k , represents ) , (
1 , 1
n m X and ) , (
1 , 2
n m X sub-band respectively,
ci w
k
X
,
1 ,
and ) , (
1 ,
n m X
ci
k
are watermarked and original DWT coefficients of chosen blocks,
i
l k ,
are
a relative measure that gives greater weight judiciously to the embedded watermark in more
salient blocks in ) , (
1 , 1
n m X and ) , (
1 , 2
n m X sub-band,
i
k 1 ,
are positive real numbers that
41
determine a tradeoff between the imperceptibility and robustness against signal distortion. The
i
k 1 ,
range between 50% and 95% of the mean value of the sub-band blocks. one PN _ is
random sequence.
Stage 3:
Perform one-level IDWT to obtain watermarked image.
4.2.2 Watermark Extracting Method:
The extraction process of watermark is rather similar to the embedding process, first we compute
DWT of the watermarked image and spilt ) , (
1 , 1
n m X and ) , (
1 , 2
n m X sub-band into non-
overlapping 8 x 8 blocks and then use the same key to generate the same random number by
which to find the watermark embedding position, and also use the same key to generate random
sequence which have the distribution of ) 1 , 0 ( N . Then we compute the correlation between
PN_one and the coefficients of selected block that embed the same watermark bit both in
) , (
1 , 1
n m X and ) , (
1 , 2
n m X sub-band and calculate the average correlation. Watermark bit value
can be decided as follows:
If correlation > 0
Watermark bit =1
else
Watermark bit =0
Watermark extraction is oblivious (blind), with no reference to the original image and thus is
more practical than non-oblivious one. The normalized correlation coefficient r was used to
measure the robustness of the extracted watermark against different attacks.
4.3 SIMULATION RESULTS AND DISCUSSION
For simulations, we take Lena image of size 512 512 as the host image shown in Fig. 4.1(a)
and watermark is visually recognizable binary image of size 32 32 shown in Fig. 4.1(b). By
using harr wavelets, we decompose Lena image into four sub-bands and watermark are
embedded in ) , (
1 , 1
n m X and ) , (
1 , 2
n m X sub-bands. We chose =90% in our simulation. . The
PSNR value of watermarked image is 37.5001 as shown in Fig. 4.1(c), and is perceptually
identical to the original host and watermark can be exactly extracted. The amplified absolute
difference between the watermarked image and host image is shown in Fig. 4.1(d). Because of
42
the adaptive and localized nature of the embedding routine the watermarks takes on
characteristics similar to the host image itself. The use of DWT and HVS allows the design of an
embedded signal which is more naturally masked by the host image itself. This permits the
embedding of a higher energy, and thus, more robust watermark. The resulting watermarked
image is corrupted using one of many common distortions which we discuss in the subsequent
section. The watermark was extracted from the corrupted image and correlated with the
embedded watermark to measure the robustness and detection capability of the technique.
The result for JPEG compression is shown in Fig. 4.2(a) for varying quality factors (QF).
Fig. 4.2 (b) and 4.2(d) shows the degraded watermarked image and Fig. 4.2 (c) and 4.2 (e) shows
the corresponding extracted watermark for quality factor 40 and 25 respectively. The result
shows that, the watermark is still present and visually detectable for quality factor of 20 and
above.
Additive white Gaussian noise was added to the watermarked image to determine the
robustness of the method to additive noise. Fig.4.3 (a) presents the result for varying SNRs. Fig.
4.3 (b), and 4.3 (d) shows the degraded watermarked image and Fig. 4.3 (c), and 4.3 (e) shows
the corresponding extracted watermark for SNR 5dB and 15dB respectively. The result shows
that, the watermark, however, had a high correlation for even high noise levels like 0dB.
The results for degradations from median filtering are also presented in Fig. 4.4(a). The
watermarked image was filtered with a F F median filter. Highly noticeable image degradation
began to occur for 5 > F . The watermark can still be detected. . Fig. 4.4 (b), and 4.4 (d) shows the
degraded watermarked image and 4.4 (c), and 4.4 (e) shows the corresponding extracted
watermark for median filtering of order 3 3 and 7 7 respectively. The results for Gaussian
low pass filtering (rotationally symmetric blur) for different standard deviation (sigma) value are
displayed in Fig. 4.5 (a). The watermark can still be detected for sigma value of above 1. Fig. 4.5
(b), and 4.5 (d) shows the degraded watermarked image and 4.5 (c), and 4.5 (e) shows the
corresponding extracted watermark for Gaussian low pass filtering for sigma value of 1 and 2
respectively.
Fig 4.6(a) shows the effect of image cropping on watermark extraction. For watermark
extraction, the portion of the watermarked image cropped out. Even when only 25% of the image
area cropped, the correlation value for the proposed technique is high about 0.85. Fig. 4.6 (b),
43
and 4.6 (d) shows the degraded watermarked image and Fig. 4.6 (c) and 4.6 (e) shows the
extracted watermark for 12.5% and 25% image area cropped respectively.




































(a)
(b)
(c) (d)
Fig.4.1. Results for Spread Spectrum-Based Watermarking Method Without any Attack: (a)
original image, (b) watermark image, (c) watermarked image, and (d) amplified absolute
difference image.
44







.




































(a)
0 10 20 30 40 50 60 70 80
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Quality Factor
c
o
r
r
e
l
a
t
i
o
n

C
o
e
f
f
i
c
i
e
n
t
s
JPEG Compression
(e)
(c)
(d)
(b)
Fig. 4.2 Results for JPEG Compression, (a)Correlation coefficient vs. Quality factor (QF), (b)
degraded watermarked image for QF=40, (c) extracted watermark for QF=40, (d) degraded
watermarked image for QF =25, (e) extracted watermark for QF=25.
45












































(a)
0 5 10 15 20 25 30 35 40 45 50
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
SNR (dB)
c
o
r
r
e
l
a
t
i
o
n

C
o
e
f
f
i
c
i
e
n
t
s
AWGN Noise
(b)
(c)
Fig.4.3 Results for Additive White Gaussian Noise Degradation, (a)Correlation coefficient vs.
SNR, (b) degraded watermarked image for SNR=5dB, (c) extracted watermark for SNR=5dB,
(d) degraded watermarked image for SNR=15dB, (e) extracted watermark for SNR=15dB.
(e) (d)
46












































(a)
2 4 6 8 10 12
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Filter Dimension
c
o
r
r
e
l
a
t
i
o
n

C
o
e
f
f
i
c
i
e
n
t
s
Median Filtering
(c) (b)
(d)
(e)
Fig.4.4 Results for Median Filtering, (a)Correlation coefficient vs. dimension of filter F, (b)
degraded watermarked image for F=3, (c) extracted watermark for F=3, (d) degraded
watermarked image for F =7, (e) extracted watermark for F=7.

47












































(a)
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
0.8
0.82
0.84
0.86
0.88
0.9
0.92
0.94
0.96
0.98
1
Filter parameter sigma
c
o
r
r
e
l
a
t
i
o
n

C
o
e
f
f
i
c
i
e
n
t
s
Gaussian low pass filtering
(d) (e)
(c)
(b)
Fig.4.5 Results for Gaussian low pass Filtering, (a)Correlation coefficient vs. filter parameter
sigma, (b) degraded watermarked image for sigma=1, (c) extracted watermark for sigma=1,
(d) degraded watermarked image for sigma =2, (e) extracted watermark for sigma=2.

48












































(a)
0 5 10 15 20 25
0.86
0.88
0.9
0.92
0.94
0.96
0.98
1
cropped image area
c
o
r
r
e
l
a
t
i
o
n

C
o
e
f
f
i
c
i
e
n
t
s
Cropping
(e)
(c) (b)
(d)
Fig.4.6 Results for Cropping, (a) Correlation coefficient vs. percent cropped image area, (b)
degraded watermarked image for 12.5% image area cropped, (c) extracted watermark for
12.5% image area cropped, (d) degraded watermarked image for 25% image area cropped, (e)
extracted watermark for 25% image area cropped.
49



The above described spread spectrum watermarking technique is a blind technique, i.e.
host signal is not required for watermark extraction, hence the method is more practical.
Simulation result shows that, the method performs better against attacks like AWGN noise,
median and Gaussian low pass filtering, cropping, and JPEG compression but experimentally,
we found that the method is not robust against mean filtering. From experience with other host
images, we find that the method works significantly better for images with highly varying
localized characteristics) i.e., images with both smooth and busy areas). This is due to the fact
that our HVS-based merging rule adapts the watermark signal strength to the local masking
characteristics of the host image. Thus, a higher energy signal can be imperceptibly embedded
with in all regions of the signal.
4.4 CHAPTER SUMMARY:
A spread spectrum based watermarking technique employing a model of the HVS was described
in the chapter. The key features of the approach are summarized below:
The host image is first decomposed using a 1
st
-level DWT. The watermark bits are
adaptively embedded through a PN-sequence in the mid-frequency sub-bands using a
HVS model to achieve good trade-off between robustness and imperceptibility.
Watermark is extracted by judging the correlation value between original PN-sequence
and coefficients of selected blocks where watermark is inserted in watermarked image.
The method is more practical as no original image is required for watermark extraction.



CHAPTER 5








CONCLUSION

Conclusion
Future Work

50
5.1 CONCLUSION
The work in this thesis, primarily focus on to provide good tradeoff between perceptual
quality of the watermarked image and its robustness to different attacks. For this purpose, we
have discussed two digital watermarking algorithms in discrete wavelet domain (DWT) by
incorporating contrast sensitivity based human visual system model (HVS). One is fusion based
watermarking, and other is spread spectrum based watermarking. We used grayscale watermark
for fusion based watermarking, and binary watermark for spread spectrum based watermarking.
Through computer simulation, we analyzed the performance of the algorithms against different
attacks such as JPEG compression, AWGN noise, mean and median filtering, cropping, and
image resizing. The important points to conclude from the simulation analysis for fusion based
watermarking algorithm were:
Embedded watermark undergoes at worst the same level of perceptible distortion as the
watermarked image.
It is resilient to JPEG lossy compression up to quality factor 5. Severe visual image
degradation occurred for quality factor of 15 and above, but still extracted watermark is
visually recognizable and correlation coefficient is high about 0.8.
It survives additive white Gaussian noise (AWGN) up to SNR of 10 dB.
It is robust to both mean and median filtering up to filter order of 13. Highly Image
degradation occurred for filter order of 9 and above, but still watermark is extractable
with correlation coefficient value of about 0.75.
It is very much robust to intentional attack cropping. Even though watermarked image is
25% cropped, the watermark is still extractable with correlation value of 0.93.
It is immune to image resizing (scaling down).
For spread spectrum based watermarking, we concluded some of important points were:
It is resilient to JPEG lossy compression up to quality factor 20.
It is very much robust to AWGN noise. Even though watermarked image is degraded by
0 dB noise, the watermark is extractable with correlation value of about 0.75.
It survives median filtering up to filter order 7, and Gaussian low pass filtering up to
sigma value 2, but its performance is not acceptable for mean filtering.
51
It is very much robust against intentional attack cropping.
This method is more practical as no original image is required for watermark extraction.
From the simulation analysis, we conclude that the both the methods are robust against different
non geometric attacks. However, both the methods fail for non-geometric attacks such as rotation
or affine transformations.
5.2 FUTURE WORK
The discussed watermarking algorithms are robust to non-geometrics attacks only. We can
extend this work by developing new watermarking algorithms, which are robust to both
geometric attacks and non geometric attacks. Future work will also concentrate on making the
watermarking methods more practical by modifying the techniques such that the host image is
not required to extract the watermark and robust to both geometric and non geometric attacks.









52
REFERENCES

[1] Van Schyndel, R.G., Tirkel, A.Z., and Osborne, C.F., A digital Watermark. Proc. of the
IEEE Int. Conference on Image Processing. Vol. 2, (1994): pp. 86-90.
[2] Swanson, M.D., Kobayashi, M., and Tewfik, A.H., Multimedia Data-Embedding and
Watermarking Technologies. Proc. of the IEEE. Vol. 86, No. 6, (June 1998): pp. 1064
1087.
[3] Petitcolas, F., Anderson, R., and Kuhn, M., Information Hiding - a Survey. Proc. of the
IEEE. Vol. 87, No. 7, (July 1999): pp. 10621078.
[4] Barni, M., Bartolini, F., Cox, I.J., Hernandez, J., and Perez-Gonzalez, F., Digital
Watermarking for Copyright Protection: A communications perspective. IEEE
Communications Magazine. Vol. 39, No. 8, (August 2001):pp. 90133.
[5] Langelaar, Gerhard C., Setyawan, I., and Lagendijk, R.L., Watermarking Digital Image
and Video Data: A state-of-the-art-overview. IEEE Signal Processing Magazine. Vol.
17, No. 5, (September 2000): pp. 20-47.
[6] Voyatzis, G., Mikolaides, N., and Pitas, I., Digital watermarking: An overview. Proc.
of IX European Signal Processing Conference(EUSIPCO), Island of Rhodes, Greece.
(September 8-11, 1998): pp. 13-16.
[7] Wolfgang, R.B., Podilchuk, C.I., and Edward J. Delp, Perceptual Watermarks for Image
and Video. Proc. of the IEEE. Vol. 87, No. 7, (July 1998): pp. 1109-1126.
[8] Cox, I.J., Miller, M.L., and Bloom, J.A., Watermarking Applications and their
Properties. Proc. of IEEE Int. Conference on Information Technology, Las Vegas.
(March 2000): pp. 6-10.
[9] Craver, S., Memon, N., Yeo, B.-L., and Yeung, M.M., Resolving Rightful Ownerships
with Invisible Watermarking Techniques: Limitations, Attacks and Implications. IEEE
Journal On Selected Areas in Communications. Vol. 16, No. 4, (May 1998): pp. 573-586.
[10] Voloshynovskiy S. et al., Attacks on Digital Watermarks: Classification, Estimation-
Based Attacks, and Benchmarks. IEEE Communication Magazine. Vol. 39. No. 8,
(August 2001): pp. 118-126.
53
[11] Gordy, J.D., and Bruton, L.T., Performance Evaluation of Digital Audio Watermarking
Algorithm. Proc .of 43
rd
IEEE Midwest Symposium on Circuits and Systems. Vol. 1,
(August 2000): pp 456-459.
[12] Ruanaidh, J.J.K.O, Dowling, W.J., and Boland, F.M., Phase Watermarking of Digital
Images. Proc. of IEEE Int. Conference on Image Processing, Lausanne, Switzerland.
Vol. 3, (September 16-19, 1996): pp. 239-242.
[13] Ruanaidh, J.J.K.O, and Pun, T., Rotation, Scale and Translation Invariant Digital Image
Watermarking. Proc. of IEEE Int. Conference on Image Processing, Santa Barbara, CA,
USA. Vol. 1, (October 1997): pp. 536-539.
[14] Cox, I.J., Kilian, J., Leighton, F.T., and Shamoon, T., Secure Spread Spectrum
Watermarking for Multimedia. Proc. of IEEE Int. Conference on Image Processing. Vol.
6, (December 1997): pp. 1673-1687.
[15] Boland, F.M., Ruanaidh, J. J. K. O, and Dautzenberg, C. Watermarking Digital Images
for Copyright Protection. Proc. of IEEE Int. Conference on Image Processing and its
Application, Edinburgh, U.K. (July 1995): pp. 321-326.
[16] Barni, M., Bartolini, F., Cappellini, V., and Piva, A., A DCT Domain System for Robust
Image Watermarking. Signal Processing Archive. Vol. 66, No. 3, (May 1998): pp. 357-
372.
[17] Burgett, S., Koch, E., and Zhao, J., Copyright Labeling of Digitized Image Data. IEEE
Communication Magazine. Vol. 36, (March 1998): pp. 94-100.
[18] Bors, A.G., and Pitas, I., Image Watermarking Using DCT Domain Constraints. Proc.
of IEEE Int. Conference on Image Processing, Lausanne, Switzerland. Vol. 3,
(September 1619, 1996): pp. 231-234.
[19] Swanson, M.D., Zhu B., and Tewfik, A.H., Transparent Robust Image Watermarking.''
Proc. of IEEE Int. Conference on Image Processing. Vol. 3, (1997): pp. 211-214.
[20] Tao, B., and Dickinson, B., Adaptive Watermarking in the DCT Domain. Proc. of
IEEE Int. Conference on Acoustics, Speech and Signal Processing, Munich, Germany.
Vol. 4, (1997): pp. 2985-2988.
[21] Podilchuk, C.I., and Zeng, W., Perceptual Watermarking of Still Images. IEEE
Workshop on Multimedia Signal Processing, Princeton, New Jersey. (June 23-25, 1997):
pp. 363-368.
54
[22] Wu, J., and Xie, J., Adaptive Image Watermarking Scheme Based on HVS and Fuzzy
Clustering Theory. Proc. of IEEE int. Conference on Neural Network and Signal
Processing, Nanjing, China. (December 14-17, 2003): pp. 1493-1496.
[23] Zhang, W., Zhu, W., and Fu, Y., An Adaptive Digital Watermarking Approach. Proc.
of IEEE int. Conference on Mechatronics and Automation, Chengdu, China. (August
2004): pp. 690-695.
[24] Pu, Y., et el., A Public Adaptive Watermark Algorithm for Color Images Based on
Principal Component Analysis of Generalized Hebb. Proc. of IEEE int. Conference on
Information Acquisition. (2004): pp. 690-695.
[25] Xia, X.-G., Boncelet, C.G., and Arce, G.R., A Multiresolution Watermark for Digital
Images. Proc. of IEEE Int. Conference. on Image Processing, Santa Barbara, CA, USA.
Vol.3, (October 26-29, 1997): pp. 548-551.
[26] Podilchuk, C.I., and Zeng, W., Image Adaptive Watermarking Using Visual Models.
IEEE Journal on Selected Areas in Communication. Vol. 16, No. 4, (May 1998): pp. 525-
539.
[27] Hsieh, M.-S., Tseng D.-C., and Huang Y.-H., Hiding Digital Watermarks Using
Multiresolution Wavelet Transform. IEEE Transaction on Industrial Electronics. Vol.
48, No. 5, (October 2001): pp.875-882.
[28] Kundur, D., and Hatzinakos D., Digital Watermarking Using Multiresolution Wavelet
Decomposition. Proc. of IEEE Int. Conference on Acoustics, Speech and Signal
Processing. Vol. 5, (1998): pp. 2969-2972.
[29] Barni, M., Bartolini F., and Piva, A., Improved Wavelet-Based Watermarking Through
Pixel-wise Masking. IEEE Transaction on Image Processing. Vol. 10, No. 5, (May
2001): pp. 783-791.
[30] Kang X., Huang J., Shi Y.Q., and Lin Y., A DWT-DFT Composite Watermarking
Scheme Robust to Both Affine Transform and JPEG Compression. IEEE Transaction on
Circuits and Systems for Video Technology. Vol. 13, No. 8, (August 2003): pp. 776-786.
[31] Wu, J., and Xie, J., Blind Wavelet-Based Watermarking Scheme Using Fuzzy
Clustering Theory. Proc. of IEEE int. Conference on Neural Network and Signal
Processing, Nanjing, China. (December 14-17, 2003): pp. 1521-1524.
55
[32] Guannan, Z., Shuxun, W., and Quan, W., An Adaptive Block-Based Blind
Watermarking Algorithm. Proc. of IEEE int. Conference on Signal Processing. (2004):
pp. 2294-2297.
[33] Mallat, S.G., Multifrequency Channel Decompositions of Images and Wavelet Models.
IEEE Transaction on Acoustics, Speech and Signal Processing. Vol. 37, No.
12,(December 1989): pp. 2091-2110.
[34] Wilson, T.A., Rogers S.K., and Myers L.R., Perceptual-based hyperspectral image
fusion using multiresolutional analysis. Optical Engineering. Vol. 34, (November 1995):
pp. 3154-3164.
[35] Levine, M. D. Vision in Man and Machine. New York: McGraw-Hill, Toronto, 1985.
[36] Kundur, D., and Hatzinakos D., Toward Robust Logo Watermarking Using
Multiresolution Image Fusion Principles. IEEE Transaction on Multimedia. Vol. 6, No.
1, (February 2004): pp. 185-197.