Digital watermarking: algorithms and applications - PowerPoint PPT Presentation


PPT – Digital watermarking: algorithms and applications PowerPoint presentation | free to download - id: 22f9c-YzU1Y


The Adobe Flash plugin is needed to view this content

Get the plugin now

View by Category
About This Presentation

Digital watermarking: algorithms and applications


S: original host signal (image luminance values or DCT coefficients) ... Watermarking of facial animation parameters (FAP) defined by the MPEG-4 standard ... – PowerPoint PPT presentation

Number of Views:3277
Avg rating:3.0/5.0
Slides: 37
Provided by: jungji8
Learn more at:


Write a Comment
User Comments (0)
Transcript and Presenter's Notes

Title: Digital watermarking: algorithms and applications

Digital watermarking algorithms and applications
  • Park, Jungjin

  • Watermarking embedding
  • Watermarking detection
  • Document
  • Graphic
  • Audio
  • Video
  • Image

Watermark embedding
Watermark embedding scheme ? embed the watermark
directly into the host data or to a transformed
version of the host data (DCT, wavelet)-popular
due to the natural framework for incorporating
perceptual knowledge into the embedding
algorithm -Many of compression techniques such as
JPEG work in the same framework and this allows
for watermarking of the compressed bit stream
with only partial decoding
Watermark embedding
S original host signal (image luminance values
or DCT coefficients) M watermark message (serial
number or credit card number logo) K secret key
Watermark embedding
? Secret key is used to generate a random
sequence to embed in the host signal ? It is also
used to determine a random sequence which
identifies locations in the host signal for
watermarking embedding -without knowledge of the
key, it should be difficult to remove or alter
the embedded message without destroying the
original content ? no-key or public-key (QIM) may
be desirable .
Watermark detection
Detection or verification the process of making
a binary decision at the decoder, it check
whether a specific watermark is or not present
in the received data ? Type I (false positive)
the case where a watermark is detected when it
does not exist ? Type 2 (false negative)
the case when a existing watermark is not
detected Identification the process of
being able to decode one of N possible choices
at the receiver. ? Open set the possibility
that one of N or no watermark exists in the
data ? Closed set the problems where one of N
possible watermarks is known to be in the
received data and the detector has to pick the
most likely one
Watermark detection
Blind detection ?S is not available at the
decoder, ?S acts as an additive noise component
in the watermarking detection process ?S is
available at the decoder ?It could be used to
estimate the channel distortions and invert them
to provide better detection performance
Watermark detection
Typical watermarking detector
Watermark detection is performed by comparing the
correlation coefficient to a threshold value
which can be modified according to the tradeoff
between probability of detection and the
probability of false alarm
Watermark W detected Watermark W is not detected
Documents watermarking
Document watermarking can be achieved by altering
the text formatting or by altering certain
characteristics of textual elements Line- Shift
Coding ?The most easily discernible by
readers ?The most robust type of encoding in the
presence of noise ?the long lengths of text lines
provide a relatively easily detectable feature
?Altering a document by vertically shifting the
locations of text lines ?decoding without need
of the original image ?Original image is known to
have uniform line spacing
Documents watermarking
Word-Shift Coding ? Altering a document by
horizontally shifting the locations of words
within text lines ? The spacing between adjacent
words on a line is often varied to support text
justification. ?less discernible to the reader
than line-shifting ? Decoding need the original
image ?variable spacing
Documents watermarking
Feature coding ? Chosen text features are altered
by extending or shorting the lengths by one or
more pixels ? Decoding require the original
image Ex) vertical end line top of letters,
b,d,h,etc Altered by expending or shorting
Graphics watermarking
Watermarking of facial animation parameters (FAP)
defined by the MPEG-4 standard ? 66 FAPs ? global
head motion parameters - Head pitch and yaw
angles ? local face motion parameters -opening of
eyelids , opening of lips, movement of innerlip
corners 16 FAPs (jaw, chin, inner lips and
cornerlips) 12 FAPs (eyeballs pupils eyelids), 8
FAPs eyebrows , 4 FAPs cheeks , 5 FAPs tongue ,
3 FAPs global head rotation, 10 FAPS outer lip
position, 4 FAPs nose, 4 FAPs ears
Graphics watermarking
Embedding ? One bit of watermark information is
embedded in a block of facial animation
parameters (FAPs) -using PN sequence ?generated
by any random number generator that produces
binary output values -1 and 1) ? Minimize
visible distortion -apply an amplitude
adaptation Limit the maximum deviation of the
watermarked FAPs from the unwatermarked FAPs to
3 of dynamic range for local FAPs like lip
movement, and 1 of the dynamic range for global
FAPs like head rotation.
Graphics watermarking
Detection ? Extracted from the watermarked
parameters directly by ?Subtraction of the
unwatermarked FAPs from the watermarked
FAPs ?Subsequent correlation with the same
filtered PN sequence that has been used for
embedding ?Thresholding as a bit decision
Video watermarking
Current issue ? Design of an effective copy
control system for DVD include s the placement of
the detector Two proposals for detector
placement ? Watermark detection in the
drive ?Advantage Pirated content cannot leave
the drive in playback mode or recording mode ?
Watermark detection within the application ?Advant
age ability to provide a more complex detector
and flexibility of extending the scheme to other
data type
Video watermarking
Unique requirement for DVD application ? Copy
generation management Ability to detect the copy
once state and change it to copy no more state
after the recording ?Two approach Secondary
watermarks, Ticket
Video watermarking
Scene-adaptive video watermarking technique ?
based on temporal wavelet transform ? using a
tow-band perfect reconstruction filter
bank ?Separates static areas from dynamic areas
so that separate watermarking strategies can be
applied to the different areas. ? constant
watermarking apply for static, varying watermark
apply for the dynamic areas to defeat watermark
deletion through frame averaging
Video watermarking
Real time watermark embedding of compressed
video ? adding the watermark by modifying the
fixed length and variable length codes in the
compressed video bit stream ? allow for a
computationally efficient way of real-time
watermark insertion ? allow for a relatively high
payload ?drawback decoding the bit stream
removes the watermark
Video watermarking
? More robust technique for real time watermark
embedding ?adding the watermark by enforcing
energy differences between various video
regions ? This technique is done by discarding
high frequency components ?only partial
decoding of a compressed video bit stream is
necessary to apply this watermark
Audio watermarking
Audio watermarking requirements ? Inaudible ?
Robust filtering, resampling, compression,
noise, cropping, A/D-D/A conversion ? Embedded
directly in the data ? self-clocking for ease of
detection in the presence of cropping and
time-scale change operations
Audio watermarking
Phase coding Work by substituting the phase of an
initial audio segment with a reference phase that
represents the data For the decoding process The
synchronization of the sequence is done before
decoding The length of the segment and the data
interval must be known at the receiver
The value of the phase of segment is detected as
a binary string
Audio watermarking
Spread spectrum ? Direct Sequence Spread Spectrum
encoding(DSSS) ?spreads the signal by multiplying
it by a chip(key), a maximal length pseudorandom
sequence- applied to the coded information to
modulate the sequence into a spread spectrum
sequence ?The spectrum of the data is spread over
the available band ?the spread data sequence is
attenuated and added to the original file as
additive random noise ?decoder ?pseudorandom
key(chip) is needed to decode ?signal
synchronization is done
Audio watermarking
?Unlike phase coding, DSSS introduced additive
random noise to the sound ?to keep the noise
level low, inaudible ?The spread code is
attenuated to roughly 0.5 percent of the dynamic
range of the host sound file
Audio watermarking
Echo data hiding ? Embedding data into a host
audio signal by introducing an echo -data are
hidden by varying three parameters of the
echo Initial amplitude, offset, decay rate
Zero represent a binary zero ,one represent a
binary one resolve the echo) ? It is possible to
encode and decode information in the form of
binary digits into a media stream with minimal
alteration to the original signal ?to minimize
alteration ? Addition of resonance simply gives
the signal a slightly richer sound
Image watermarking
Embed m-sequences into the least significant bit
(LSB) of the data ? provide an effective
transparent embedding technique ? good
correlation properties (for detection) ?
computationally inexpensive to implement Texture
block coding ? Hide data within the continuous
random texture patterns of a picture ?
Implemented by copying a region from a random
texture pattern found in a picture to an area
that has similar texture
Image watermarking
  • Texture block coding Detection
  • Autocorrelate the image with itself. This will
    produce peaks at every point in the
    autocorrelation where identical regions of the
    image overlap.
  • 2. Shift the image as indicated by the peaks in
    Step 1.
  • Now subtract the image from its shifted copy
  • 3. Square the result and threshold it to recover
    only those values quite close to zero. The copied
    region will be visible as these values.

Image watermarking
Transform domain watermarking ? robust to common
compression techniques ? block-based DCT which is
the fundamental building block of current image
coding standard JPEG and MPEG ? a pseudorandom
subset of the blocks are chosen and a triplet of
midrange frequencies are altered to encode a
binary sequence ? Watermarks inserted in the high
frequencies are vulnerable to attack ? The low
frequency components are perceptually significant
and sensitive to alterations
Image watermarking
?two watermarking techniques based on visual
models ? Image-adaptive DCT approach ?
Image-adaptive DWT approach ? Utilizing visual
models which have been developed in the context
of image compression ? Very effective visual
models have been developed for compression
applications that take into account frequency
sensitivity, local luminance sensitivity,
contrast masking
Image watermarking
Visual model ?frequency sensitivity human eyes
sensitivity to sine wave gratings at various
frequencies ?depend on the modulation transfer
function of the eye and is dependent of the image
data ?Luminance sensitivity measure the effect
of the detectability threshold of noise on a
constant background ?Contrast masking the
detectability of one signal in the presence of
another signal and the effect is strongest when
both signals are of the same spatial frequency ,
orientation, and location ?combination of the
components JND thresholds for the entire image
Image watermarking
IA-DCT embedding
DCT coefficient Watermarked DCT
coefficients Sequence of watermark
values Computed JND from the visual model
?The watermark is only inserted into the
luminance component of the image
Image watermarking
Detection Normalized correlation detection scheme
based on classical detection for the IA-DCT scheme
Received watermark
Normalized correlation coefficient between two
Image watermarking
IA-W embedding
Wavelet coefficient at position (u,v) in
resolution level l, frequency orientation
f Watermarked wavelet coefficient Computed
frequency weight at level l and frequency
orientation f Watermark sequence
? watermark is inserted only in the luminance
component of the image
Image watermarking
IA-W detection Correlation is performed
IA-W scheme is based on a much simpler visual
model which only takes into account frequency
sensitivity, the multi resolution structure of
the watermark and the watermark detection scheme
results in a very robust scheme
Image watermarking
Digital watermarking by geometric warping Embeds
information in an image by changing the geometric
features of the image ?the watermark is formed by
a predefined dense pixel pattern, such as a
collection of lines ?Salient points in an image
are warped into the vicinity of the line pattern
in such a way that the changes to the image are
imperceptible ? subdivide the image in a number
of blocks. Find a fixed number of most
significant pixels, these are called salient
Image watermarking
Detection ?Determining whether a significantly
large number of points are within the vicinity of
the line patterns Advantage ?detection is
computationally faster ?Easier to detect the
watermark in images have been rotated, scaled, or
distorted by a geometric transformation