Perceptual Redundancy

Here is an image represented with 8-bits per pixel

 Perceptual Redundancy
The same image at 7-bits per pixel
 Perceptual Redundancy
At 6-bits per pixel
 Perceptual Redundancy
At 5-bits per pixel
 Perceptual Redundancy
At 4-bits per pixel
 Perceptual Redundancy
It is clear that we don’t all these bits!
Our previous example illustrated the eye’s sensitivity
to luminance
We can build a perceptual model
Give more importance to what is perceivable to the
Human Visual System
Usually this is a function of the spatial frequency
Fundamentals of JPEG

Encoder

DCT Quantizer Entropy coder

Compressed
image data

IDCT Dequantizer Entropy

decoder

Decoder
 Entropy
• In information theory, entropy is a measure of the
uncertainty in a random variable. The term usually refers to
the Shannon entropy, which quantifies the expected value
of the information contained in a message.
• A fair coin has an entropy of one bit. A series of two fair
coin tosses has an entropy of two bits. The number of fair
coin tosses is its entropy in bits.

• However, if the coin is not fair, then the uncertainty, and

hence the entropy rate, is lower. This is because, if asked to
predict the next outcome, we could choose the most
frequent result and be right more often than wrong. 8
 Fundamentals of JPEG
JPEG works on 8×8 blocks
Extract 8×8 block of pixels
Convert to DCT domain
Quantize each coefficient
Different stepsize for each coefficient
Based on sensitivity of human visual system
Order coefficients in zig-zag order
Similar frequencies are grouped together
Run-length encode the quantized values and
then use Huffman coding on what is left
 MPEG
MPEG – Moving Pictures Experts Group
Coding of moving pictures and associated audio
Picture part
Can achieve compression ratio of about 50:1 through storing only the difference
between successive frames
Even higher compression ratios possible
 Image Compression
Because much of this information is graphical or pictorial in
nature, the storage and communications requirements are
immense.

Image compression addresses the problem of reducing the

amount of data requirements to represent a digital image.

Image Compression is becoming an enabling technology:

HDTV.

Also it plays an important role in Video Conferencing, remote

sensing, satellite TV, FAX, document and medical imaging.
 Run- Length Coding
Run-length encoding (RLE) is a very simple form
of data compression in which runs of data (that is,
sequences in which the same data value occurs in
many consecutive data elements) are stored as a
single data value and count, rather than as the original
run. This is most useful on data that contains many
such runs.
WWWWWWWWWWWWBWWWWWWWWWW
WWBBBWWWWWWWWWWWWWWWWWW
WWWWWWBWWWWWWWWWWWWWW
12W1B12W3B24W1B14W
12
 Bit-Plane Coding
An effective technique to reduce inter pixel redundancy is to
process each bit plane individually

The image is decomposed into a series of binary images.

Each binary image is compressed using one of well known

binary compression techniques.
 Bit Plane Coding
A bit plane digital discrete signal (such as
image or sound) is a set of bits corresponding
to a given bit position in each of the binary
numbers representing the signal.
For example; 16-bit data representation there
are 16 bit planes: the first bit plane contains
the set of the most significant bit, and the 16th
contains the least significant bit.

14
 Bit Plane Coding
It is possible to see that the first bit plane gives
the roughest but the most critical
approximation of values of a medium, and the
higher the number of the bit plane, the less is
its contribution to the final stage. Thus, adding
a bit plane gives a better approximation.
If a bit on the nth bit plane on an m-bit dataset
is set to 1, it contributes a value of 2(m-n),
otherwise it contributes nothing.
15
For example, in the 8-bit value 10110101 (181 in decimal) the
bit planes work as follows:
Bit Plane Value Contribution Running
Total
1st 1 1 * 2^7 = 128 128
2nd 0 0 * 2^6 =0 128
3rd 1 1 * 2^5 = 32 160
4th 1 1 * 2^4 = 16 176
5th 0 0 * 2^3 = 0 176
6th 1 1 * 2^2 = 4 180
7th 0 0 * 2^1 = 0 180
16
8th 1 1 * 2^0 = 1 181
17
 Transform Coding
A reversible linear transform (such as Fourier
Transform) is used to map the image into a set of
transform coefficients
These coefficients are then quantized and coded.
The goal of transform coding is to decorrelate pixels
and pack as much information into small number of
transform coefficients.
Compression is achieved during quantization not
during the transform step
Transform Coding

Software Research
Wavelet Coding
 Wavelet Coding
High Frequency coefficients tend to be very
small --- 0

They can be quantized very effectively without

distorting the results
 Wavelet Compression
Wavelet compression is a form of data
compression well suited for image compression
(sometimes also video compression and audio
compression). Notable implementations are JPEG
2000, DjVu and ECW for still images, REDCODE
and CineForm for video.
The goal is to store image data in as little space as
possible in a file. Wavelet compression can be either
lossless or lossy.
22
 Wavelet Compression
Using a wavelet transform, the wavelet compression
methods are adequate for representing transients,
such as percussion sounds in audio, or high-
frequency components in two-dimensional images,
for example an image of stars on a night sky. This
means that the transient elements of a data signal
can be represented by a smaller amount of
information than would be the case if some other
transform, such as the more widespread discrete
cosine transform, had been used.
23
Visible & Invisible Watermarking

Dr. Engr. Junaid Zafar

24
 Outline

Introduction 
State-of-the-Art 
Characteristics of Visible & Invisible 
Watermarking Schemes
Attacking Visible & Invisible Watermarking 
Schemes
Discussions and Conclusions 
Classifying Watermarking Schemes

Data hiding

Steganography Watermarking

Imperceptible Visible Imperceptible Visible

data embedding data embedding watermarking watermarking

Non-robust Robust Fragile Robust

data embedding data embedding watermarking watermarking
Visible Watermarking

+ 

Unobtrusive copyright patterns can be recognized •

on embedded contents.
Invisible Watermarking v.s.
Visible Watermarking
Invisible Watermarking Visible Watermarking

Watermark
Imperceptible distortion Visibly Meaningful
Perceptibility
pattern

Intentional attacks and User-intervention

Robustness
common signal based
processing watermark removal
Extraction Explicit extraction Direct viewing
module
Current
Hot Only few papers
Research Status
Requirements of
Visible Watermarking
Perceptibility of host image details 
Contents should not be rendered useless after being 
visibly watermarked
Visibility of watermark patterns in embedded 
contents
No explicit watermark extraction techniques are 
required
Robustness 
Difficult to remove unless exhaustive and expensive 
human interventions are involved
A General Model of Visible Watermarking

I '  K 1 * I  K 2 *W
D ( E I ( I ' ), E I ( I ))  Threshold I
D ( EW ( I ' ), EW (W ))  Threshold W
I’: the watermarked content 
I: the un-watermarked original content 
W: the watermark pattern 
Ki: the weighting factor 
D: a distance function measuring the perceptual difference of its two 
parameters
Ei: image feature extraction operators 
ThresholdI: the largest allowable distortion of image details that observers 
can tolerate and, at the same time, the signature of can be maintained.
ThresholdW: the largest allowable distortion of the embedded watermark 
pattern that the copyright information can be clearly recognized.
 Other Enhancing Schemes
[Meng and Chang] 
The same embedding model is extended to the DCT domain by simple 
statistic model approximation for the convenience of processing directly
in the MPEG-compressed domain.
[Kankanhalli et al] 
Local features related to the degree of distortion tolerances, such as 
edge locations, texture distributions and luminance sensitivity, are taken
into consideration so that more unobtrusive watermarked images can be
generated.
Simple statistics of block-DCT coefficients are calculated and analyzed 
to decide the watermark embedding energy of each block.
Edge integrity will be preserved, in these approaches, since the 
edge information is essential to maintain the image quality.
And the energy of the embedded watermark is larger in highly 
textured areas than in smooth ones due to different noise
sensitivity.
In additions, the watermark energy of mid-gray regions is also 
smaller than other areas since the noises are more visible against a
mid-gray background
[S. P. Mohanty et al] 
in addition to the visibly embedded watermark, a fragile invisible 
watermark is also adopted to check if the visible watermark is altered or
Important observations (1/4)

Attacking visible watermarking scheme 

means successfully recover the
watermarked area.
Implication: 
Similar image processing techniques can be 
adopted
Image recovery 
Object removal 
Important observations(2/4)

To clearly recognize the copyright 

patterns, the contours of embedded
patterns must be preserved.
Implication: 
An attacking scheme is effective if 
The pattern is completely removed .1
The shape is seriously distorted without seriously .2
degrading visual quality.
Important observations(3/4)
The perceptibility of the host image details 
within watermarked area depends on the
preservation of edge information.
Implication: 
Available information while attacking
Surrounding pixels around watermarked area. 
Edge information within watermarked area is 
available while attacking.
Important observations(4/4)

The robustness lies in the inevitability of 

exhaustive and expensive labors.
Implication: 
Only minimum user intervention should be 
adopted during attacking
User selection of watermarked areas 
Averaging Attacks

Refill the watermarked areas by averaging 

surrounding pixels.
Good approximations for small areas. 
Blurring effects across object boundaries 
Image Inpainting
M. Bertalmio, V. Caselles, and C. Ballester, 
“Image inpainting,” SIGGRAPH 2000, Aug.
2000
 n N (i , j , n )  n
I (i , j )   L (i , j ) 
t
n
  I (i , j )

 N (i , j , n ) 

Image inpainting 
is an iterative image recovery technique. 
successfully reconstruct the edges of damaged area. 
Basic Inpainting Attacks

Attacks against visible watermarking are regarded as common 

image recovery problems.
Good results can be obtained for areas composed of thin copyright 
patterns, but areas composed of thick patterns cannot be
successfully recovered.
 General Attacks
Watermark Area

Edge Area

Classifying flat areas within watermarked area 

by analyzing remaining edge information of host
images
Directly extend colors of surrounding flat areas 
into watermarked areas
Further Improvement
Experimental Results (I)
Experimental Results (II)
Experimental Results (III)
Outline
What is Watermark?
Usage
Previous approaches
Difficulties and problems
Algorithms to digital watermarking using
wavelets.
We’ll focus on the wavelet’s contributions
to the algorithms.
conclusions
44
Electronic publishing

Inexpensive copies
No quality loss
Wide distribution but no control

Now can we protect copyright ???

45
The use of digital watermarks

Can be used as an authentication tool.

As a method to discourage the
unauthorized copying and distribution of
electronic documents.

(Data Hiding )

46
What is watermarking?

Copyright protection methods.

Consists of signing an image with a
signature or copyright message.
The message is secretly embedded in the
image.
There is no visible difference.

47
Requirements
Requirements vary with application. For 
example:
Perceptually transparent - must not 
perceptually degrade original content.
Robust - survive accidental or malicious 
attempts at removal.
Oblivious or Non-oblivious - Recoverable 
with or without access to original.
Capacity – Number of watermark bits 
embedded.
Efficient encoding and/or decoding.
Example: Additive Watermarks

I(x,y) IW (x,y)

Multiply by gain
k factor k IW(x,y)= I(x,y)+ k  W(x,y)

W (x,y): Pseud o Rand om Pattern {-1,0,1}

Additive watermarks in transform
domain: Spread Spectrum

fv

fh
Watermark Attacks

Active Attacks. 
Hacker attempts to remove or destroy the 
watermark.
Watermark detector unable to detect 
watermark.
Key issue in proof of ownership, fingerprinting, 
copy control.
Not serious for authentication or covert 
communication.
Covert Communication

A covert channel is a type of computer

security attack that creates a capability to
transfer information objects between
processes that are not supposed to be
allowed to communicate by the computer
security policy.
A covert channel is so called because it is
hidden from the access control mechanisms
of ultra-high-assurance secure operating
systems. 52
Covert channels are exceedingly hard to 
install in real systems, and can often be
detected by monitoring system
performance.
In addition, they suffer from a low signalto- 
noise ratio and low data rates (on the
order of a few bits per second). They can
also be removed manually with a high
degree of assurance.
53
Watermark Attacks
Passive Attacks. 
Hacker tries to find if a watermark is present.
Removal of watermark is not an aim.
Serious for covert communications.
Collusion Attacks. 
Hacker uses several copies of watermarked 
data (images, video etc.) to construct a copy
with no watermark.
Uses several copies to find the watermark.
Serious for fingerprinting applications.
Watermark Attacks

Forgery Attacks. 
Hacker tries to embed a valid watermark.
Serious in authentication.
If hacker embeds a valid authentication 
watermark, watermark detector can accept
bogus or modified media.
Example: Forging Photo I.D.
Millionaire
with Swiss
bank Bad
account Forgery
!

Thanks
Hong Man,
for this
picture!

Assistant
Professor
Content-based Watermarking

Original Watermark
JPEG Compression Attack
 Watermarking Research
Information Theoretic Issues. 
Compute “watermarking capacity”
How many watermark bits can be hidden in 
a watermark channel ?
Capacity when multiple watermarks are 
used.
Decision Theoretic Issues. 
Robust watermark detector.
Quickest possible watermark detection.
Invariant to statistical uncertainties.
Achieves desired false alarm and detection 
probabilities.
 Watermarking Research
Signal Processing Issues. 
What transforms to use? – wavelet, DCT etc.
Geometric invariant transform – scale, 
translation, and rotation invariance.
Watermarking protocols and system 
issues.
Protocol used by sender and receiver for 
secure transmission.
Steganalysis. 
How to analyze multimedia data for the 
presence of watermark(s).
 Robust Watermarking Techniques -
A Communications Framework

Media Carrier Media Carrier

Coding and Decoding and

Water-
X +
Modulation Demodulation Watermark
mark

Noise/Attack Channel
Information Theoretic Formulation
A new watermark channel model. 
Multiplicative and additive attacks.
Models geometric attacks (“killer” attack).
Attack parameters assumed imperfectly 
known.
Lower and upper bounds on watermarking 
capacity.
Many current capacity estimates are specific 
cases of proposed estimates.
Capacity in the presence of multiple 
watermarks using multiuser information theory.
Multiple Description Watermarking

A new framework for watermarking using 

multiple description codes.
Partition the source into multiple descriptions.
Each description carries enough information.
Allocate watermarking bits to descriptions 
subject to distortion constraints.
Experiments for secure wireless media 
transmission is promising.
Rate-distortion theory formulation. 
Decision Theoretic Formulation

How many pixels to watermark ? 

A new measure for watermarking capacity.
Different from standard information theoretic 
formulation.
Perhaps a more practical measure.
Fast watermark detector. 
Computation reduction.
Useful for hardware implementation.
Steganalysis

A distributed detection framework for 

analyzing images for presence of
watermarks.
Theoretical formulation and practical 
algorithm for steganalysis of LSB based
image watermarking.
A new watermarking capacity measure 
called “steganographic capacity”.
Projected Applications of Proposed
Techniques

Secure media streaming. 

Wireless multimedia security. 
Watermark error resilience coding. 
Error concealment for wireless multimedia 
transmission.
Difficulties
 Digital watermarking algorithms usually use the
lower-order bit-planes of the original image, so
do intentional disturbance algorithms.
 Cannot be inserted to downgrade the quality of
the source image too much.
 Digital watermark readers are usually widely
available.
 There is a limited amount of data that can be
used to insert digital watermarks in a highly
compressed JPEG image;
 Noticeable artifacts of image compression
usually destroy watermarks easily.
67
Motive of watermarking

Recent advancements in computer 

technologies offer many facilities for
duplication, distribution, creation, and
manipulation of digital contents.
Encryption is useful for transmission but 
does not provide a way to examine the
original data in its protected form.
Watermarking process
Embedding stage (1
Spatial domain 
flipping the low-order bit of each pixels 
Frequency domain 
embedding the watermark in mid-frequency components 
relatively robust to noise, image processing and 
compression
the quality of the host image will be distorted significantly if 
too much data is embedded
Distribution stage (2
Compression, transmission error, and common 
image processing are seen as an attack on the
embedded information
Spatial Domain Embedding

70
Watermarking process
Extraction stage (3
Blind – extraction without original image 
Semi-blind – rely on some data or features 
Non-blind – need original image 
Detection stage (4
Evaluate the similarity between the original and 
detected watermark
False positive：watermark is detected although there is 
none
False negative：no watermark is detected while there is 
one
Watermarking properties
Perceptual transparency 
Robustness 
The mark should resist to 
Common signal processing like lossy compression 
Geometric transformation like image rotation, scaling, and 
cropping
Security 
How easy it is to intentionally remove a watermark 
Data capacity 
Amount of information that can be stored within the 
content
Categories of digital watermark
Perceptible 
Imperceptible 
Robust 
ownership assertion 
Fragile 
indicate modifications of the content 
Semi-fragile 
differentiate between lossy transformation that are 
“info. preserving” and lossy transformation which are
“info. altering”
Applications
Copyright Protection 
Invisible watermark which can tolerate malicious and 
unintentional attacks
It does not prevent people from copying the digital 
data
Data Hiding 
It tries to invisibly embed the maximum amount of 
data into a host signal => this allows communication
using enciphered messages without attracting the
attention of a third party
Robustness is not important while invisibility and 
capacity are required
Applications
Authentication and Data Integrity 
Verification watermarks are required to be fragile, so 
that any modification to the image will destroy the
mark
Copy Protection 
Requirements 
Robustness against removal 
Ability of blind detection 
Capability of conveying non-trivial number of bits 
Examples of watermarking

Image Watermarking 
Document Watermarking 
Graphics watermarking 
Video watermarking 
Image Watermarking
Robust and imperceptible 
The watermark may be scaled appropriately to 
minimize noticeable distortion to the host
Examples 
Texture-based watermark 
Embed it into a portion of the image with similar texture 
Insert a watermark into the phase components 
The phase information is perceptually more important than 
the magnitude data
Document Watermarking
Line-shifts 
Word-shifts 
Document Watermarking
Slight modifications to characters 
Graphics Watermarking
Embedding in facial animation parameter (FAP) 
data
The amount of deviation the watermark signal has on 
FAP is limited to minimize visible distortion
1% for head rotation 
3% for lip motion 
Video Watermarking
Copy generation management 
Minimum information that must convey 
Copy never 
Copy once 
Copy no more 
Copy freely 
For DVD, real-time decoding is required 
Detector placement 
Detection in the drive 
Detection within the application (within MPEG 
decoder)
 Video Watermarking
Examples 
For video coding like MPEG or H.26x, we embed the 
watermark into DCT coefficients
Only partial decoding of block DCT is necessary for watermark 
embedding
If constant bit rate is required, only nonzero DCT coefficients are 
marked
Layered Access Control Schemes
on Watermarked Scalable Media
Scalability – spatial, temporal and SNR 
scalability
A typical MPEG-2 conditional access receiver 
Layered Access Control Schemes
on Watermarked Scalable Media
A typical MPEG-2 conditional access receiver 
Control Word – change very often so that fast (simple) 
decryption algorithm would not be broken easily
EMM (Entitlement Management Message) – with the 
aid of EMM, the broadcaster can change the status of
the user accessibility of contents
ECM (Entitlement Control Message) 
Layered Access Control Schemes
on Watermarked Scalable Media
A typical MPEG-2 conditional access receiver 
Problem 
We send the key to user via reliable channel, while the 
content goes through an unreliable channel.
Key and contents won’t arrive at the same time 
=> Synchronization problem
Solution 
Using robust watermark to embed the ECM and EMM 
information into the content
Existing methods
Method in the spatial domain and in the
frequency domain.
Visible watermarking systems are usually
able to sustain all possible image
alterations and even intentional
disturbances.

However, image quality is significantly

reduced.
86
Existing methods
 Recent frequency domain water-marking
methods are based on the discrete cosine trans-
form (DCT),where pseudo-random
sequences,such as M-sequences, are added to
the DCT coefficients at the middle frequencies
as signatures.
 This approach,of course,matches the current
image/video compression standards well,such
as JPEG,MPEG 1-2,etc.

87
 Fingerprinting or traitor tracing
Marking copies of one document with a customer signature.

original
+
…
W1 W2 WN

… N customers
Robust, secure, invisible watermark, resistant with respect
to the collusion attack (averaging copies of documents with
different marks).
 Adding captions to images, additional
information to videos
Typical application:
Adding subtitles in multiple languages•
Additional audio tracks to video•
Tracking the use of the data (history file)•
Adding comments, captions to images•

Watermark requirements:
Moderately robust scheme•
Robustness with respect to lossy compression, noise adding,•
and A/D D/A conversion
Original images (frames) not available for message extraction•
Security requirement not so strong •
Fast detection, watermark embedding can be more time •
consuming
 Watermarking principles
In spatial domain
watermark embedded
by directly modifying
the pixel values
In transform domain
watermark embedded in the
transform space by modifying
coefficients

+ =
DCT Inverse DCT

Modify
DCT
Watermarking for color images
One or more selected color channels. •
Luminance•

Oblivious vs. non-oblivious watermarking

non-oblivious = original image is needed for extraction
oblivious = original image is not necessary
 Robustness to geometric transformations
Easy if the original image is available (non-oblivious schemes)

Very challenging for oblivious schemes especially for a

combination of cropping, scaling, rotation, and shift

Approaches:
Watermarking by small blocks (good for cropping)•
Embedding patterns with known geometry•
Watermarking using Fourier-Mellin transform (scaling and •
rotation converted to shift)
Embedding watermarks into image features or salient points•

Weak points:
Computational complexity•
More powerful geometric attacks - StirMark•
 Forensic analysis

Analysis of lighting and shadows

Localized analysis of
- noise
- histogram
- colors

Looking for discontinuities

 Fragile watermarks

Break easily
Properties: Computationally cheap
Good localization properties
Too sensitive for redundant data

Examples:
Embedding check-sums in the LSBs
Adding m-sequences to image blocks
 Robust watermarks on small blocks

Medium robustness
Insensitive to small changes
Properties:
Not as good localization properties
Can distinguish malicious and
non-malicious modifications

Spread spectrum watermarks on

Examples: medium size blocks
Wavelet domain watermarks
1. Insert robust watermark into every block

64 pixels

Robust
bit extractor
B 50 bits
Watermarked
Secret W(K, B) B block B
key K Synthesizing
Gaussian + =
Block # sequence
B
 Hybrid watermark

Fragile, sensitive, and robust

Properties: Good localization properties
Can distinguish malicious and
non-malicious modifications

Examples:
Robust watermarks on medium blocks
combined with a fragile watermark
 Images with Selfcorrecting Capabilities

Content of block B1 is compressed and encoded in the•

LSBs of B2
B1 and B2 are separated by a random vector p•
 Reconstruction of a license plate

Tampered image - The license plate The original license plate after
has been replaced with a different one reconstruction

2 LSBs have been used for selfembedding•

Reconstruction after mosaic filtering

Manipulated image Reconstructed image

Secret key
 Oblivious secure watermarking

State of the art: Robustness with respect to changes in

gray levels and simple geometric transformations such as
shift, scaling, rotation, and cropping

Needs to be solved: Robust watermark with a computationally

efficient detector that can extract watermarks from images that
underwent a combination of gray level mapping and general
geometric distortions

Possible approaches:
Content Locked Coordinate Systems •
Feature-based techniques•
Problems with DCT
 A given image cannot be queried for
ownership without the original un-
watermarked image (non- oblivious).
 Robustness.
 It is known that the wavelet image/video coding,
are included in the image/video compression
standards, such as JPEG 2000 and MPEG4 due
to excellent performance in compression.
 Wavelet transforms set high frequency
components = 0 (Transients eliminated).
 Therefore, it is important to study water-marking
methods in the wavelet transform domain. 103
Watermarking in the DWT domain
DWT has a key advantage it has over
Fourier transforms is temporal resolution: it
captures both frequency and location
information (location in time).
includes two parts:
encoding
Adding the watermark to the original image.
decoding.
Recognizing or extracting the watermark.

104
The DWT

Furthermore, from these DWT coefficients,

the original signal can be reconstructed.

This process is called the inverse DWT

(IDWT).

105
 DWT pyramid decomposition
 An image can be
decomposed into a
pyramid structure with
various band information
.
 such as: HH,LH ,LL and
HL frequency bands.

106
Encoding and decoding scheme
1. De-compose an image into several bands with
a pyramid structure.
2. Add the watermark message.
3. Then, we take the two dimensional IDWT of
the modified DWT coefficients.
4. The decoding will be done by applying the
inverse procedure.

107
What we are going to see?

 Three different methods:

 Adding Pseudo-random codes to the high and
middle frequencies- Delware U.
 Adding the mark with respect to the human
visual system -Toronto U.
 Extracting the mark without the original host
Wave- Mark- Stanford U.
 For each will see its advantage, and
examples.
108
First method - Delware U.

 Multi-resolution watermarking method for

digital images.
 Wavelet transform based watermarking.
 pseudo-random codes to the large
coefficients at the high and middle
frequency bands of the DWT of an image
 The message is secretly embedded in
the image and there is no visible
difference between the original and the
signed one. 109
First method -Encoding
1. Calculating the DWT coefficients y[m,n]
2. The message is a Gaussian noise N[m,n]: with
mean 0 and variance 1.

  control the level of watermarking

 Squre2 indicates amplifications of the large
coefficients.
 We do not change the DWT coefficients at the
lowest resolution!
110
Encoding

3. we take the two dimensional IDWT of

the:
 modified DWT coefficients
 unchanged DWT coefficients at the lowest
resolution.

111
Encoding-
4. For the resultant image to fit within the 0
to 255 integer values:

5. This is resultant watermark image.

112
First method –Decoding
DWT HL1
-
LH1 HH1 cross
HH1
Watermarked Image correlation
with the
watermark
DWT HL1

is there a
LH1 HH1
HH1 peak?
Original Image

YES NO

113
STOP
First method –Decoding
DWT HL1
-
LH1 HH1 cross
Watermarked Image correlation
with the
watermark
DWT HL1

is there a
LH1 HH1
peak?
Original Image

YES NO

114
STOP
First method -Decoding

6. Otherwise,we consider the signature

added in the HLI,LH1,and HHI
bands,we continue to decompose the
original and the received signals in the
LL1 band into four additional subbands
LL2,LH2,HL2 and HH2 and so on until a
peak appears in the cross correlations.

7. Otherwise,the signature can not be

115
detected.
First method - Advantages

 The method is hierarchical.The computation

load needed to detect the watermark depends
on the noise level in an image.

 Adding watermarks on these large coefficients

is difficult for the human eyes to perceive.

 Matches the emerging image/video

compression standards. Robust to wavelet
transform based image compressions scheme,
and as well as to other common image
distortions. 116
Second method
 Stage I: The host image and the
watermark are transformed into the
wavelet domain.
 Only the 1st level discrete wavelet
decomposition of the watermark is
performed.
 We perform the Lth level discrete
wavelet decomposition of the host
image.
117
Second method
 Size of the watermark: Nwx X Nwy
 Stage II: The detail images of the host at each
resolution level are segmented into non-
overlapping Nwx X Nwy rectangles.
 The watermark is embedded by a simple
scaled addition of the watermark to the
particular Nwx X Nwy.
 The scaling of the watermark is a function of
the salience of the region. The greater the
salience S, the stronger the presence of the
watermark.
118
Second method

 Stage III: The corresponding Lth level

inverse wavelet reconstruction of the
fused image components is performed to
form the watermarked image.
 Mathematically, contrast sensitivity is
defined as the reciprocal of the contrast
necessary for a given spatial frequency
to be perceived.

119
Second method- decoding

 The watermark is extracted, 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.
120
The Human Visual System
 Multi-resolution wavelet decomposition of both
the host image and the watermark.
 When an image undergoes a wavelet
decomposition, its components are separated
into bands of approximately equal bandwidth
on a logarithmic scale much as the retina of
the eye splits an image into several
components.
 It is, therefore, expected that use of the DWT
will allow the independent processing of the
resulting components much like the human
eye. 121
Second method advantages
1. Provides a simultaneous spatial localization
and frequency spread of the watermark within
the host image.
2. In addition, the watermark merging process is
adaptive as it depends on the local image
characteristics at each resolution level.
3. Robust as it embeds the watermark more
strongly into more salient components of the
image.

122
Example

123
The third method-overview
 The algorithm in WaveMark uses discrete
wavelet transforms and error-correcting
coding schemes to provide robust
watermarking of digital images.
 The watermark recovery procedure does not
require a match with an uncorrupted original
image.
 The system is practical for real-world
applications, encoding or decoding images at
the speed of less than one second each on a
Pentium Pro PC.
124
Using wavelet
 The use of Daubechies' advanced
wavelets makes the watermarked
images more perceptively faithful than
Haar.
 The watermark is adaptively applied to:
 Different frequency bands
 Different areas,based on the smoothness
 increases robustness within the limits of
perception.

125
Encoding

 We 1st convert and store the image in a

component color space with intensity and
perceived contrasts.

 For each color component, we perform a

4-level wavelet transform using
Daubechies-4 wavelet.

126
Encoding

127
Error correction coding
A Hamming code is used to add
redundancy to the bits so that the errors
can be detected or corrected to a certain
extent.
we use a (8,4) extended Hamming code.
4 bytes are the input
8 bytes include the input +4 parity bytes.
Such a code can correct a one-bit error
and detect up to three one-bit errors.
128
Getting the final result
 After the watermark code is placed in the
transform matrices, we perform a 4-level inverse
wavelet transform for each of the three matrices
using Daubechies-4 wavelet.

 Then we use the inverse color transformation to

obtain a coded color image in the RGB color
space.

 This image is the final image coded with the

duplicated watermark codes.
129
Results
fast speed
 high reliability
Robustness- alterations including
compression, intentional disturbances and
image processing operations.
 The algorithm is able to detect the
invisible digital watermark within each
altered image after performing Hamming
error-correcting decoding.
130
Limitations
Wavelet Watermarking is designed to
sustain image alterations such as
compression, additive noise and even
some intentional disturbances.

 However, like other wavelet-based

watermarking algorithms, it is not suited to
handle significant rescaling, aspect ratio
changes and rotational transformations.
131