Get documents about "Imaging Watermarking
Shared by: yaoyufang
-
Stats
- views:
- 2
- posted:
- 10/9/2011
- language:
- English
- pages:
- 28
Document Sample
Imaging Watermarking
Survey and Ongoing Current Research
Ahmed Abu-Hajar, Ph.D.
Digitavid, Inc.
San Jose, CA
Presentation Outline
Review of Digital Image Processing
Introduction to Image Watermarking
Spatial Domain Watermarking Techniques
Transform Domain Watermarking Techniques
Our proposed Transform Technique
Results
Conclusion
Review of Digital Image Processing
An image is a 2-D Signal X (Rows)
Pixel
– Spatial signal
– Intensity value as I(X,Y)
Y (Cols)
Digital image
– Digitized 2-D signal
Use rectangular shape areas
called pixels
Digital representation: 8-bit, Lena Image
12-bit, 16-bit, …
Review of Digital Image Processing
Digital image are represented
using matrices
I(1,1) I(1,2) I(1, N)
I(x, y) = I(2,1) I(2,2) I(2, N)
I(M,1) I(M,2) I(M, N) Lena Image
Review of Digital Image Processing
Digital Image Processing
Filtering
– FIR, IIR
– Low pass, High pass,…
Down sampling/ up sampling
Transformation: DFT, DCT, Wavelet, …
Compression, modulation,…
Review of Digital Image Processing
Lena Image
Weak LPF
Strong LPF Weak HPF
Introduction to Watermarking
What is watermarking?
– Watermarking is embedding a hidden message within
the original data “host image”
Why watermarking is used?
– Proof of Ownership ( copyrights and IP protection)
– Copying Prevention
– Broadcast Monitoring
– Authentication
– Data Hiding
Introduction to Watermarking
Image watermarking became popular in the 1990s
because of the widespread of the Internet
I believe controlling the Internet is a losing battle
– It is just like betting one million dollars on me,
To win a marathon
– It would never happen
– Even if I practiced very hard
In my opinion, watermarking is useful when the
number users is limited.
Introduction to Watermarking
Problem Statement
– A hidden watermark message is inserted into a host image
such that the hidden message will survive intended or
unintended attacks
Host Watermarked Attacked
Image Image Image
Watermark Watermark Yes/No
Attack
I(x,y) Insertion IM(x,y) Detection
W(x,y))
Watermark Watermark
Message Message
M(x,y) M(x,y)
Introduction to Watermarking
Watermark Insertion
– The process of adding the watermark message
Host Watermarked Attacked
Image Image Image
Watermark Watermark Yes/No
Attack
I(x,y) Insertion IM(x,y) Detection
W(x,y))
Watermark Watermark
Message Message
M(x,y) M(x,y)
Types Watermarking
Watermark message M(x,y)
– Random or pseudo random signal
– Binary { -1,+1} or { -1, 0, +1}
– other signals are used
Watermark message is added linearly as:
W ( x, y ) = I ( x, y ) + kM ( x, y )
I(x,y) W(x,y)
kM(x,y)
Types Watermarking
The hidden message may be added in
– Spatial Domain
– Discrete Fourier Transform (DFT) Domain
– Discrete Cosine Transform (DCT) Domain
– Discrete Wavelet Transform Domain (DWT)
– Fractals Domain
– Other Transforms Domains
Spatial Domain Watermarking
The watermark message is added in the spatial
domain
I M ( x, y ) = I ( x, y ) + kM ( x, y )
I(x,y) IM(x,y)
M(x,y) Watermarked
Host
Image Image
Watermark Message
Spatial Domain Watermarking
The watermarked image undergo an attack such as compression.
The watermark message is detected using correlation coefficient
between the watermark message and the attacked image.
∑ W ( x, y )M ( x, y )
x,y
ρ =
∑ W 2 ( x, y )
x,y
∑x,y
M 2 ( x, y )
Attacking
IM(x,y) Image Yes/No
Watermark
Attack
Detection
W(x,y))
Watermarked
Image
Spatial Domain Watermarking
The correlation coefficient is compared against some threshold value T
The theoretical bounds of this approach are based on the spread
spectrum technique and limited by the channel’s capacity
ρ > T YES
ρ ≤ T NO
Attacking
IM(x,y) Image Yes/No
Watermark
Attack
Detection
W(x,y))
Watermarked
Image
Spatial Domain Watermarking
Different types of attacks are considered
– Compression
– Scaling (resizing)
– Filtering ( low pass, high pass, …
– Adding noisy signal
Attacking
IM(x,y) Image Yes/No
Watermark
Attack
Detection
W(x,y))
Watermarked
Image
Transform Domain Watermarking
The watermark message is inserted in the transform domain
Different transforms behaves differently to different attacks
May support human visual system (HVS)
I(x,y)
T(.) T-1(.)
M(x,y)
Host Image Watermarked
Image
Watermark Message
Transform Domain Watermarking
Discrete Fourier Transform (DFT) is superior for shifting
attacks
– Shifting in the space domain leads to a phase shift in the
frequency domain.
I(x,y)
T(.) T-1(.)
M(x,y)
Host Image Watermarked
Image
Watermark Message
Transform Domain Watermarking
Discrete Cosine Transform (DCT) 10 20 15 20
20 40
– Supports block-based transform (8x8), x=
22 25
18 20 49 50
(16x16), (32x32), …
50 28 40 30
– Compacts the energy of each block
114.2500 -18.9956 4.7500 7.0564
according to its frequency content -31.1702 -5.7730 -3.6587 -16.2760
X =
– Used in JPEG compression -7.7500 23.6342 -1.2500 0.2225
-1.4306 -13.7760 -6.8731 3.7730
– The watermark message is embedded
in the intermediate frequency
coefficients
4x4 block 4x4 DCT
block
Transform Domain Watermarking
Discrete Cosine Transform
– Example of adding watermark image to the intermediate
frequency of each block
Watermarked Watermark
Image message
Transform Domain Watermarking
Discrete Wavelet Transform DWT
– DWT locally separates the content
of the image into low frequency and
high frequency subbands.
– Most of the energy is concentrated
in the low frequency subband.
– In watermarking: The message is
inserted in the high frequency
subbands (HL,LH and HH)
Vast number of embedding
techniques already developed
Correlation coefficients are also
based on CDMA techniques
Transform Domain Watermarking
Non-Regular Transform
– Spreads the energy content into different subbands
– The subbands are similar to the original image
Similar images
Similar histogram
Transform Domain Watermarking
Our proposed watermarking scheme is based on Non-
Regular Transform
– It inserts the watermark in the transform domain such that the
message is more resilient to attacks.
– The message contains small frequency and high frequency
contents.
Correlation
Image
T(.) T-1(.) T(.) ∫
Image Message
Watermark Attack
Insertion Detection
Transform Domain Watermarking
Our Results
– The average correlation coefficient for all the subbands after
JPEG compression at different Q
0.7
0.6
0.5
Correlation Coeff.
GRS4
0.4
Daub4
daub8
0.3
bior9/7
0.2
0.1
0
90 70 50 30 10
Q
Transform Domain Watermarking
Our Results
– The average correlation coefficient for all the subbands after
JPEG2000 compression at different bit rates
0.8
0.7 GRS4
daub4
0.6
daub8
0.5 biorth9/7
0.4
0.3
0.2
0.1
0
2 1 0.5 0.25
Bit Rate
Transform Domain Watermarking
Our Results
– The average correlation coefficient for all the subbands after
JPEG compression at qualities Blind Detection
0.35
0.3
0.25
R
G S4
0.2
Daub4
daub8
0.15
bior9/7
0.1
0.05
0
90 70 50 30 10
Q
Transform Domain Watermarking
Our Results
– The average correlation coefficient for all the subbands after
JPEG2000 compression at bitrates Blind Detection
0.35
0.3
0.25
GRS4
0.2 duab4
0.15 daub8
biorth9/7
0.1
0.05
0
2 1 0.5 0.25
Bit Rate
Conclusion
Review of Digital Image Processing
Introduction to Image Watermarking
Spatial Domain Watermarking Techniques
Transform Domain Watermarking
Techniques
Our proposed Transform Technique
Results
