I. Cox et al., “Secure spread spectrum watermarking for
multimedia,” IEEE Transactions on Image Processing, 6(12):
1673-1687, 1997.
Dianbo GONG
31008039
DIGITAL WATERMARKING
BACKGROUND：WHY WE NEED WATERMARK
Conventional cryptographic systems permit only valid key holders
access to encrypted data, but once such data is decrypted
there is no way to track its reproduction or retransmission.
Therefore, conventional cryptographic provides little protection
against data piracy, in which a publisher is confronted with
unauthorized reproduction of information. A digital watermark
is intended to complement cryptographic processes.
WATERMARK
It is a visible, or invisible,
identification code that is
permanently embedded in
the data and remains
present within the data after
any decryption process.
EXAMPLE OF A DIGITAL WATERMARK
A simple example of a digital watermark would be a
visible “seal” placed over an image to identify the copyright
owner.
watermark
visible “seal” placed over an image
VISIBLE WATERMARK
A visible watermark is limited in many ways.
1.It lower the image fidelity
2. A visible watermark is susceptible to be attacked through
direct image processing.
In order to be effective, a watermark should have the
characteristics outlined below.
Unobtrusiveness: The watermark should be invisible, or its
presence should not interfere with the work being protected.
Robustness: The watermark must be difficult to remove.
Universality: The same digital watermarking algorithm
should apply to all three media such like audio, image and
video data under consideration.
Unambiguousness: Retrieval of the watermark should
clearly identify the owner.
ROBUSTNESS
In particular, the watermark should be robust in the following areas.
Common signal processing:
The watermark should still be retrievable even if common signal
processing operations are applied to the data. These include, digital-toanalog and analog-to-digital conversion, and resampling.
Common geometric distortions:
Watermarks in image and video data should also be immune from
geometric image operations such as rotation, translation, cropping and
scaling.
PREVIOUS DIGITAL WATERMARKING METHODS
Several previous digital watermarking methods have
been proposed.
But those methods still have problems such like：these
methods may not be robust to common geometric
distortions such like cropping.
many compression schemes (JPEG,MPEG,) can potentially
degrade the data’s quality.
Cropping, or the cutting out and removal portions of an image,
leads to irretrievable loss of image data, which may seriously
degrade any spatially-based watermark .
However, a frequency-based scheme spreads the watermark over
the whole spatial extent of the image, therefore less likely to be
affected by cropping. So it’s better than previous digital
watermarking methods we mention.
HOW TO BUILDING A STRONG WATERMARK
There are two parts to building a strong
watermark:
1.The watermark structure
2.The insertion strategy(where and how).
In order for a watermark to be robust and secure,
these two components must be designed correctly.
In this paper author provide two key insights that make
the watermark both robust and secure:
1.The watermark should be composed of random
numbers from a Gaussian distribution（independent
identically distribution）
2.The author argue that the watermark should be
placed in the perceptually most significant components
of the data.
The author discovered that the perceptually significant components
have a capacity that allows watermark insertion without degradation.
Further, most processing techniques applied to media data tend to leave
the perceptually significant components intact. So the watermark can still
remain after those processing.
For that reason we should focus on the perceptually significant spectral
components of the data.
Once the perceptually significant components are located, Gaussian
noise is injected. The choice of this distribution gives resilient
performance against attacks.
DC COMPONENTS
AC LOW-FREQUENCY COMPONENTS
AC HIGH-FREQUENCY COMPONENTS
AC HIGH-FREQUENCY COMPONENTS
A watermark placed in the high-frequency spectrum of an image
can be easily eliminated with little degradation to the image by any
process that directly or indirectly performs like low pass filtering.
In fact, data loss usually occurs among the high-frequency
components.
An scaling (shrinking) of the image leads to a loss of data in the
high-frequency spectral regions of the image.
reason why don’t use high-frequency spectral regions
CHOOSE AC LOW-FREQUENCY：
Reason choose AC Low-frequency：
First, the location of the watermark is not obvious.
Second, A watermark that is well placed in the Low-frequency
domain of an image will be practically impossible to see.
Further, spectrum analysis of images reveals that most of the
information in such data is located in the low-frequency regions.
PROBLEM: HOW TO INSERT A WATERMARK
To solve this problem, the frequency domain of the image
is viewed as a communication channel.
The watermark is viewed as a signal that is transmitted
through it.
Attacks and unintentional signal distortions are treated as
noise that the watermark must be immune to.
CONCEPT: SPREAD SPECTRUM COMMUNICATION
In spread spectrum communications: one transmits a narrowband
signal over a much larger bandwidth such that the signal energy
present in any single frequency is undetectable.
In our case, the watermark is spread over very many frequency bins
so that the energy in any one bin is very small and certainly
undetectable.
Because the watermark verification process knows the location and
content of the watermark, it is possible to concentrate
these many weak signals into a single output with high signal-to-noise
ratio (SNR).
DCT（DISCRETE COSINE TRANSFORMATION）
In principle, any frequency domain transform can be used.
in this paper the author use a Fourier domain method based on
the DCT.
DCT in matlab code：
dct(),dct2():
idct(),idct2(): inverse DCT
59
 61
F 
62

59
60 58 57 
59 59 57 
59 60 58

61 60 56
DCT
IDCT
 236.25 4.5169
 1.0592  0.1768
D
  1.25  0.4387

 0.7093  0.2803
 1.75
1.1056 
1.1713  0.7803
 2.25  1.7125 

0.8678 0.1768 
INSERT A WATERMARK
In order to place a length n watermark into an N*N image, we
computed the DCT of the image and placed the watermark into the
highest magnitude coefficients of the transform matrix, excluding the
DC component. For most images, these coefficients will be the ones
corresponding to the low-frequencies.
By using Si=Xi (1+αWi)
Xi ：originally matrix
Wi： watermark
α： A ratio number that control the watermark Robustness
INSERTING THE WATERMARK
59
 61
F 
62

59
60 58 57 
59 59 57 
59 60 58

61 60 56
60

~  61
F
62

59
60 58 57 
59 59 57 
59 60 58

61 60 55
DCT
IDCT
IDCT
 236.25 4.5169
 1.0592  0.1768
D
  1.25  0.4387

 0.7093  0.2803
 1.75
1.1056 
1.1713  0.7803
 2.25  1.7125 

0.8678 0.1768 
Insert
watermark
 1.75 1.1056 
236.25 5.5169
0.0592  0.1768 1.1713  0.7803

D
  1.25  0.4387  2.25  1.7125 


0
.
7093

0
.
2803
0
.
8678
0
.
1768


DETECTION OF THE WATERMARK
1.Use DCT for the original picture and
the watermark picture
2. Proceeds by adding all of these
very small signals, and concentrating
them once more into a signal output
with high signal-to-noise ratio.
DETECTION OF THE WATERMARK
 236.25 4.5169  1.75 1.1056 
 1.0592  0.1768 1.1713  0.7803

D
  1.25  0.4387  2.25  1.7125 


0
.
7093

0
.
2803
0
.
8678
0
.
1768


By using：Xi ' =[(Ai-Bi) /Bi]α
Ai： matrix that including the
watermark information
Bi：originally matrix
 1.75 1.1056 
236.25 5.5169
0.0592  0.1768 1.1713  0.7803

D
  1.25  0.4387  2.25  1.7125 


0
.
7093

0
.
2803
0
.
8678
0
.
1768


Compare these
two matrix and
extracted the
watermark
information
extracted watermark X’
α：A ratio number that control
the watermark Robustness
EVALUATING THE SIMILARITY OF WATERMARKS
extracted watermark X’
original watermark X
X'X
sim (X, X' ) 
X'X'
To decide whether X’ and X match，we setting
the detection threshold. sim（X，X’）>T
NECESSARY CHARACTERISTICS
By following EXPERIMENTAL RESULTS shows： this digital watermarking
method satisfied the necessary characteristics
1.Fidelity preservation
2.Robustness to common signal and geometric processing operations
3.Robustness to attacks.
4.Can use to audio, image and video data.
Unobtrusiveness: The watermark should be perceptually invisible,
or its presence should not interfere with the work being protected.
Robustness: The watermark must be difficult to remove.
Universality: The same digital watermarking algorithm
should apply to all three media under consideration.
Unambiguousness: Retrieval of the watermark should clearly
identify the owner.
FOR EXAMPLE: CROPPING
Only the central quarter of the image
remains.
In order to extract the watermark from
this image, the missing portions of the
image were replaced with portions from
the original unwatermarked image
X'X
sim (X, X' ) 
X'X'
Result： the response of the
watermark is still bigger than the
threshold.
REFERENCE：
I. Cox et al., “Secure spread spectrum watermarking for multimedia,”
IEEE Transactions on Image Processing, 6(12): 1673-1687, 1997.
image Watermarking in DCT：an Embedding Strategy and Algorithm
HUANG Ji-wu，2，Yun Q. SHI，CHENG Wei-dong