watermarking algorithm
-
Upload
ashish-khandelwal -
Category
Documents
-
view
234 -
download
0
Transcript of watermarking algorithm
-
7/30/2019 watermarking algorithm
1/31
INTRODUCTION:
Digital watermarking is a technique which allows an individual to add
hidden copyright notices or other verification messages to digital audio,
video, or image signals and documents. Such a message is a group of bits
describing information pertaining to the signal or to the author of the signal
(name, place, etc.). The technique takes its name from watermarking of
paper or money as a security measure. Digital watermarking can be a form
of steganography, in which data is hidden in the message without the end
user's knowledge
In other words Digital watermarking is the process of embedding
information, or a watermark, into a digital multimedia object such that
the watermark can be detected or extracted later to make an assertion
about the object. Watermarking has proven to be a reliable mean toprovide copy protection and authenticity proof for digital media, and
therefore a lot of research has been performed in these areas.
The concept of watermarking comes from more than 700 years back. It was
a technique used by paper manufacturers to identify their products. Today,
watermarks in paper can still be seen. Along with the years the concept of
watermarking has penetrated into the field of security. Currency, such as
dollar bills, checks, postal stamps, and official documents from
government can be seen to carry watermarks. Besides these paper-based
applications, watermarking can also be used to provide the same degreeof security to digital media data, such as audio, text and still images.
Digital watermarking is an adaptation of the commonly used and well
known paper watermarks to the digital world. Digital watermarking
describes methods and technologies that allow hiding information, for
example a number or text, in digital media, such as images, video and audio.
The embedding takes place by manipulating the content of the digital data
that means the information is not embedded in the frame around the data.
The hiding process has to be such that the modifications of the media are
imperceptible. For images this means that the modifications of the pixel
values have to be invisible. Furthermore, the watermark has to be robust or
fragile, depending on the application. With robustness we refer to the
capability of the watermark to resist to manipulations of the media, such as
lossy compression, scaling, and cropping, just to enumerate some. An
example of watermarking is shown below
http://en.wikipedia.org/wiki/Copyrighthttp://en.wikipedia.org/wiki/Watermarkhttp://en.wikipedia.org/wiki/Steganographyhttp://en.wikipedia.org/wiki/Copyrighthttp://en.wikipedia.org/wiki/Watermarkhttp://en.wikipedia.org/wiki/Steganography -
7/30/2019 watermarking algorithm
2/31
A watermark is a recognizable image or pattern in paper that appears lighter
when viewed by transmitted light (or darker when viewed by reflected light,
atop a dark background). A watermark is made by impressing a water-coated
metal stamp or dandy roll onto the paper during manufacturing. Watermarks
were first introduced in Bologna, Italy in 1282; they have been used by
papermakers to identify their product, and also on postage stamps, currency,
and other government documents to discourage counterfeiting.
An image with visible digital watermarking
-
7/30/2019 watermarking algorithm
3/31
Hardware implementations in digital watermarking are scarce and may
not be justified for most applications. Several hardware implementations
are intended for various applications, were software solutions tend to
struggle to satisfy the computational demands. A high percentage of these
works are ASIC-based implementations (Application Specific Integrated
Circuit). Only two of the works found are FPGA-based implementations
(Field Programmable Gate Array).
FPGA devices have improved on speed, capacity, flexibility, and power
dissipation over the years. Current applications where FPGA can be
utilized include digital signal processing, computer vision, speech
recognition, computer hardware emulation, and cryptography. Applications
that contain heavy amounts of parallelism can benefit the most from
the FPGA architecture . For video watermarking, the FPGA should
provide the benefits of parallel processing and specific-architecturedesign offered by Application Specific Integrated Circuits(ASIC), but at a
fraction of the price.
Therefore, the goals of this thesis are to select a watermarking method
appropriate for working on various images, and to develop a hardware
implementation of the watermarking algorithm for these images in an
FPGA device, in order to study the issues related to implementing
such an algorithm in an FPGA device, and explore the potential of
performing such algorithms in hardware platforms. Since the Digital
Signal Processor (DSP) is traditionally used to handle similar tasks, acomparison with the FPGA implementation is provided. The comparison
will be in terms of performance or speed, power dissipation, unit cost,
and development cost. The results will be discussed and compared to
previous work in the area.
-
7/30/2019 watermarking algorithm
4/31
Background and Review:
Watermarking is defined as the direct embedding of additional
information into the original content or host signal, and in as a
technique to embed invisible or inaudible data within multimedia
content. Usually the watermark is imperceptible to the human. There should
be no way to remove or modify the watermark without changing or altering
the content (or signal). The watermark can carry or provide information. In
general, watermarking must comply with the following three
requirements: (1) imperceptibility, (2) robustness and (3) capacity. In
[3] the authors found that sometimes these requirements conflict with
each other. The tradeoff between these requirements depends on the
application the watermark is intended for. For example, if one desires
to test for tampering, one would employ an algorithm that guaranteesimperceptibility but that is not robust to any modification of the
content (this is termed as fragile or semi-fragile watermark). On the
other hand, if one desires a copyright protection that must withstand an
irreversible or lossy transformation or additional attacks, a very robust
watermarking algorithm would then be selected. Some transformations,
or attacks, that the signal may be subject to are resampling rescaling,
compression, linear and nonlinear filtering, additive noise, A/D and D/A
conversion. Watermarks typically are not retrieved, they are only detected.
Detection is usually performed by correlation methods, correlating the
watermarked data with the watermark sequence. The value of
correlation is compared against a threshold value, which is then used
to decide if the watermark was detected or not. The threshold value is
determined by the application and trial-and-error runs.
Watermarking can have various purposes which include copyright
protection, authentication, tamper detection, and data hiding. It can be
applied to different media types such as digital images, video, graphics,
audio, text, and multimedia content. The creation of the internet and the
conversion of audio-visual and textual content to digital format haveallowed replicating and distributing digital content over and over,
without any visible penalty on the data. Therefore, there seems to be an
increasing desire to protect property rights for digital media. The authors
agree that in the past 10 years there has been a new and great interest in the
area of digital watermarking as a way to help to protect authenticity and
prevent unauthorized replication of media.
-
7/30/2019 watermarking algorithm
5/31
Watermarking is as old as paper manufacturing itself because watermarks
were a by-product of the process of making paper. Years ago our ancients,
during the process of making paper, poured a mix of slurry of fiber and
water on to mesh molds to collect the fiber. Then this slurry was dispersed to
add shape and uniformity, and finally great pressure was applied to the
mesh molds in order to expel the water and cohere the fiber [6]. Years
ago this by-product, coined watermark, was use to establish the
authenticity of a product or certify something about the product. In
present days this same principle is applied using digital watermarks. And as
the authors in [6] state, whether the product of paper press or discrete
cosine transformations, watermarks of varying degrees of visibility are
added to presentation media as a guarantee of authenticity, quality
ownership and source.
The watermarking technique has evolved from steganography, but
steganography and watermarking have their differences. In watermarking,
protecting the content that carries the watermark is essential, whereas in
steganography the content is of no value and the message that is
covered in the content is the significant one. So the applications of both
concepts are very different, but the uses sometimes overlap.
Typical application for watermarking include digital archives, copyright
protection, legal delivery of content, anti-piracy, and automatic broadcast
monitoring. One example of a practical use of watermarking comes from
the Academy Awards. When the Academy sends screeners to its voters,
the movies include a watermark in each of its frames. A distinct watermark
is used for each recipient. If the movie gets illegally distributed, the
watermark then allows the Academy to know which voter was the source
for the pirated version.
-
7/30/2019 watermarking algorithm
6/31
Applications of Watermarking:
This section describes seven applications of watermark-ing: broadcast
monitoring, owner identification, proof of ownership, authentication,
transactional watermarks, copy control and covert communication.
(i) Broadcast monitoring:
This is one of the most exclusive applications of Watermarking. In 1997, a
scandal broke out in Japan regarding television advertising. At least two
stations had been routinely overbooking air time. Advertisers were paying
for thousands of commercials that were never aired .The practice had
remained largely undetected for over twenty years, in part because therewere no systems in place to monitor the actual broadcast of advertisements.
There are several types of organizations and individuals interested in
broadcast monitoring. Advertisers, of course, want to ensure that they
receive the air time purchased from broadcasting firms. Musicians and actors
want to ensure that they receive accurate royalty payments for broadcasts of
their performances and copyright owners want to ensure that their property
is not illegally rebroadcast by pirate stations. We can use watermarks for
broadcast monitoring by putting a unique watermark in each video or sound
clip prior to broadcast. Automated monitoring stations can then receive
broadcasts and look for these watermarks, identifying when and where eachclip appears. Commercial systems have been deployed for a number of years
and the basic concepts have a long history.
(ii) Owner identification
Although a copyright notice is no longer necessary to guarantee copy rights,
it is still recommended. The form of the copyright notice is usually c_date,owner. On books and photographs, the copyright is placed in plane sight. In
movies, it is appended to the end of the credits. And on prerecorded music, itis placed on the packaging. One disadvantage of such text copyright notices
is that they can often be removed from the protected material. And images
can be spatially cropped. A digital watermark can be used to provide
complementary copyright marking functionality because it becomes an
integral part of the content, i.e. the copyright information is embedded in the
music to supplement the text notice printed on the packaging. The Digimarc
-
7/30/2019 watermarking algorithm
7/31
Corporation has marketed a watermarking system designed for this
application. Their watermark embedder and detector are bundled with
Adobes popular image processing program, Photoshop. When the detector
finds a watermark, it contacts a central database to identify the watermarks
owner (who must pay a fee to keep the information in the database).
(iii) Proof of ownership
Multimedia owners may want to use watermarks not just to identify
copyright ownership, but to actually prove ownership. To illustrate the
problem, lets quickly introduce some characters who are well known in the
watermarking literature. Suppose Alice creates an image and puts it on her
website, with a copyright notice c_Alice 2000. Bob then steals the image,
uses an image processing program to replace the copyright notice with c
_Bob 2000, and then claims to own the copyright himself. How can thedispute resolved?
Traditionally, Alice could register the image with the Copyright Office by
sending a copy to them. The Copyright Office archives the image, together
with information about the rightful owner. When the dispute between Alice
and Bob comes up, Alice contacts the Copyright Office to obtain proof that
she is the rightful owner. If Alice did not register the image, then she should
at least be able to show the film negative. However, with the rapid
acceptance of digital photography, there might never have been a negative.
In theory, it is possible for Alice to use a watermark embedded in the image
to prove that she owns it. However, this is not a trivial problem.
(IV) Authentication
As both still and video cameras increasingly embrace digital technology, the
ability for undetectable tampering also increases. The content of digital
photographs can easily be altered in such a way that it is very difficult to
detect what has been changed. In this case there is not even an original
negative to examine. There are many applications where the veracity of an
image is crucial, especially in legal cases and medical imaging.
Authentication is a well studied problem in cryptography .Friedman first
discussed its application to create a trustworthy camera by computing a
cryptographic signature that is associated with an image. If even one bit of
-
7/30/2019 watermarking algorithm
8/31
one pixel of the image is modified, it will no longer match the signature, so
any tampering can be detected. However, this signature is metadata that
must be transmitted along with the photograph, perhaps in a header field of a
particular image format. If the image is subsequently copied to another file
format that does not contain this header field, the signature will be lost, and
the image can no longer be authenticated.
A preferable solution is to embed the signature directly into the image using
watermarking. This eliminates the problem of ensuring that the signature
stays with the image. It also opens up the possibility that we can learn more
about what tampering has occurred, since any changes made to the image
will also be made to the watermark. Thus, there are several systems that can
indicate the rough location of changes that have been made to the image.
There are also systems designed to allow certain changes, such as JPEG
compression [18, 19], and only disallow more substantial changes, such asremoving an individual from a crime scene.
(v)Transactional watermarks (Fingerprinting)
Monitoring and owner identification applications place the same watermark
in all copies of the same content. However, electronic distribution of content
allows each copy distributed to be customized for each recipient. This
capability allows a unique watermark to be embedded in each individualcopy. Transactional watermarks, also called fingerprints, allow a content
owner or content distributor to identify the source of an illegal copy. This is
potentially valuable both as a deterrent to illegal use and as a technological
aid to investigation.
One possible application of transactional watermarks is in the distribution of
movie dailies. During the course of making a movie, the result of each days
photography is often distributed to a number of people involved in its
production. These dailies are highly confidential, yet occasionally, a daily is
leaked to the press. When this happens, studios quickly try to identify the
source of the leak. Clearly, if each copy of the daily contains a unique
transactional watermark that identifies the recipient, then identification of
the source of the leak is much easier.
-
7/30/2019 watermarking algorithm
9/31
Another application of transactional watermarks was deployed by the DiVX
Corporation. DiVX marketed a modified version of DVD. One of the
security measures implemented in DiVX hardware was a transactional
watermark that could be used to identify a player used for piracy. If illegal
copies of a DiVX movie turned up on the black market, DiVX could use the
watermark to track them to the source.
(VI) Copy Control
Transactional watermarks as well as watermarks for monitoring,
identification, and proof of ownership do notpreventillegal copying. Rather,
they serve as powerful deterrents and investigative tools. However, it is also
possible for recording and playback devices to react to embedded signals. In
this way, a recording device might inhibit recording of a signal if it detects a
watermark that indicates recording is prohibited. Of course, for such asystem to work, all manufactured recorders must include watermark
detection circuitry. Such systems are currently being developed for DVD
video and for digital music distribution. Interestingly, the use of watermarks
in video to control equipment dates back to at least 1989 and in audio to
perhaps 1953.
(vii) Covert communication
One of the earliest applications of watermarking, or more precisely, data
hiding, is as a method of sending secret messages. The application has beenformulated by Simmons as the prisoners problem, in which we imagine
two prisoners in separate cells trying to pass messages back and forth. Their
problem is that they cannot pass these messages directly, but rather, must
rely on the prison warden to act as a messenger. The warden is willing to
carry innocuous messages between them, but will punish them if he finds
that, for example, their messages relate to a plan for escape. The solution is
to disguise the escape-plan messages by hiding them in innocuous messages.
There are several commercially available programs designed for this
application, including Stego Tools.
-
7/30/2019 watermarking algorithm
10/31
Properties:
There are a number of properties that have discussed the characteristics of
watermarks some of the properties discussed are robustness, tamper
resistance, fidelity, computational cost, and false positive rate. In practice, it
is probably impossible to design a watermarking system that excels at all of
these. Thus, it is necessary to make tradeoffs between them, and those
tradeoffs must be chosen with careful analysis of the application. In addition,
the application can affect the very definition of a property. In the following
subsections, we look at each of the five properties listed above, and discuss
how its importance and definition varies with application.
(i) Robustness
A watermark is said to be robust if it survives common signal processing
operations such as digital-to-analog-to conversions and lossy compression.
More recently, there has been an increased concern that video and still image
watermarks also be robust to geometric transformations.
Robustness is often thought of as a single-dimensional value, but this is
incorrect. A watermark that is robust against one process may be very fragile
against another. In many applications, robustness to all possible processing
is excessive and unnecessary.
Usually, a watermark must survive common signal processing only between
the time of embedding and the time of detection. For example, in television
and radio broadcast monitoring, the watermark need only survive the
transmission process. For television, this means lossy compression, analog
transmission, and some small amount of horizontal and vertical translation.
It need not survive rotation, scaling, high-pass filtering, or any of a wide
variety of distortions that do not occur during broadcast.
In some cases, robustness may be completely irrelevant, or even undesirable.Watermarks used for covert communication need not be robust at all, if the
cover media will be transmitted digitally without compression. A watermark
For simple authentication, which just indicates whether the media has been
altered, should be fragile.
-
7/30/2019 watermarking algorithm
11/31
On the other hand, when the signal processing between embedding and
detection is unpredictable, the watermark may need to be robust to every
conceivable distortion. This is the case for owner identification, proof of
ownership, fingerprinting, and copy control. It is also true for any
application in which hackers might want to remove the watermark.
(ii)Tamper resistance
Tamper resistance refers to a watermarking systems resistance to hostile
attacks. There are several types of tamper resistance. Depending on the
application, certain types of attacks are more important than others. In fact,
there are several applications in which the watermark has no hostile
enemies, and tamper resistance is irrelevant. Some basic types of attack are
Activeattacks. Here the hacker tries to remove the watermark or make itundetectable. This type of attack is critical for many applications, including
owner identification, proof of ownership, fingerprinting, and copy control, in
which the purpose of the mark is defeated when it cannot be detected
.However; it is not a serious problem for authentication or covert
communication.
Passive attacks. In this case, the hacker is not trying to remove the
watermark, but is simply trying to determine whether a mark is present, i.e.
is trying to identify a covert communication. Most of the scenarios above are
not concerned with this type of attack. In fact, we might even advertise thepresence of the mark so that it can serve as a deterrent. But for covert
communication, our primary interest is to prevent the watermark from being
observed.
Collusionattacks. These are a special case of active attacks, in which the
hacker uses several copies of one piece of media, each with a different
watermark, to construct a copy with no watermark .Resistance to collusion
attacks can be critical in a fingerprinting application, which entails putting a
different mark in each copy of a piece of media. However, the number of
copies that we can expect the hacker to obtain varies greatly from
application to application. For example, in the DiVX application, a hacker
can buy any number of DiVX players, and play one movie on all of them to
obtain any number of differently-watermarked copies. On the other hand, in
the film-studio dailies application, each employee can only obtain one copy
-
7/30/2019 watermarking algorithm
12/31
of the watermarked material. A collusion attack would require that several
employees conspire to steal the material, which is an unlikely prospect.
Forgery attacks. Here, the hacker tries to embed a valid watermark,
rather than remove one. These are our main security concern in
authentication applications, since, if hackers can embed valid authentication
marks, they can cause the watermark detector to accept bogus or modified
media this type of attack is a serious concern in proof of ownership.
(iii) Fidelity
A watermark is said to have high fidelity if the degradation it causes is very
difficult for a viewer to perceive. However, it only needs to be imperceptible
at the time that the media is viewed. If we can be certain that the media will
be seriously degraded before it is viewed, we can rely on that degradation tohelp mask the watermark. Such a case occurs when we watermark video that
will be transmitted over NTSC, or audio that will be transmitted over AM
radio. The quality of these broadcast technologies is so low that our initial
fidelity need not be very good. Conversely, in HDTV and DVD video and
audio, the signals are very high quality, and require much higher fidelity
watermarks (though, of course, the quality of the content remains the same -
a bad movie is a bad movie whether on VHS or DVD).
In some applications, we can accept mildly perceptible watermarks in
exchange for higher robustness or lower cost. For example, Hollywooddailies are not finished products. They are usually the results of poor
transfers from film to video. Their only purpose is to show those involved in
a film production the raw material that has been shot so far. A small visible
distortion caused by a watermark will not diminish their value.
(iv) Computational cost
Different applications require the embedders and detectors to work atdifferent speeds. In broadcast monitoring, both embedders and detectors
must work in (at least) real time. The embedders must not slow down the
media production schedule, and the detectors must keep up with real-time
broadcasts. On the other hand, a detector for proof of ownership will be
valuable even if it takes days to find a watermark. Such a detector will only
be used during ownership disputes, which are rare, and its conclusion about
-
7/30/2019 watermarking algorithm
13/31
whether the watermark is present is important enough that the user will be
willing to wait.
Furthermore, different applications require different numbers of embedders
and detectors. Broadcast monitoring typically requires a few embedders and
perhaps several hundred detectors at different geographic locations. Copy
control applications may need only a handful of embedders but millions of
detectors. Conversely, in the fingerprinting application implemented by
DiVX, in which each player embeds a distinct watermark, there would be
millions of embedders and only a handful of detectors. In general, the more
numerous a device needs to be for a given application, the less it must cost.
The wide variation in dollar cost and in speed requirements means that there
is a wide variation in the required computational efficiency of watermark
embedders and detectors.
(v) False positive rate
A false positive is a detection of a watermark in a piece of media that does
not actually contain that watermark. When we talk of the false positive rate,
we refer to the number of false positives we expect to occur in a given
number of runs of the detector. Equivalently, we can discuss the probability
that a false positive will occur in any given detector run. There are two
subtly different ways to define this probability that are often confused in the
literature. They differ in whether the watermark or the media is consideredto be the random variable.
In the first definition, the probability of a false positive is the probability
that, given a fixed piece of media and a randomly-selected watermark, the
detector will report that the watermark is in the media. The watermarks are
drawn from a distribution that is defined by the design of a watermark
generation system. Typically, watermarks are generated by either a bit-
encoding algorithm or by a Gaussian, independent random number
generator. In many cases, probability of false positives, according to this
first definition, is actually independent of the piece of media, and depends
only on the method of watermark generation.
In the second definition, the probability of a false positive is the probability
that, given a fixed watermarkand a randomly-selected piece of media, the
detector will detect the watermark in the media. The media is chosen from
the distribution of natural media, which is defined by either nature or
-
7/30/2019 watermarking algorithm
14/31
Hollywood, depending on the application. This distribution is very different
from that defined by the watermark generation system, and thus probabilities
based on this definition can be quite different from those based on the first
definition.
In most applications, we are more interested in the second definition of false
positive probability than in the first. However, in a few cases, the first
definition is also important, such as in the case of fingerprinting, where the
detection of a random watermark in a given image might lead to a false
accusation of theft.
The probability of false positives that is required depends on the application.
In the case of proof of ownership, the detector is used so rarely that a
probability should suffice to make false positives unheard of. On the other
hand, in the copy control application, millions of watermark detectors areconstantly being run on millions of pieces of media all over the world. If one
piece of unwatermarked media consistently generates false positives, it
could cause serious trouble. For this reason, the false positive rate should be
in-finitesimal. For example, the general consensus is that watermark
detectors for DVD video should have a false positive rate of 1 in 1012frames.
Types of Watermark:
In we have a good decomposition of the variety of watermarks
currently available, their definitions, their features and possible
applications, advantages and disadvantages . One thing to point out of
this classification is that video watermarking is an extension of image
watermarking, which utilizes characteristics of the Human Visual
System (HVS) to embed the watermark. HVS methods take advantage of
the way the human eye processes images in order to add watermarks to the
images. The video method requires that the watermarking encoding or
decoding process be done in real-time and that the watermark is robustfor compression, that is, the watermark should be present in the signal
after compressing the video. According to the survey done in
Podilchuk et al. were the first to state that for the watermark to be robust
it had to be embedded into the perceptually significant portions of the
data, although Cox et al. were also pioneers in watermarking perceptually
significant areas.
-
7/30/2019 watermarking algorithm
15/31
Watermark processing methods fall into two categories: spatial domain and
frequency domain. Spatial domain usually changes the value of the
pixels in a minor way so it is not perceived by the human eye and the
watermark is scattered through the entire object. Frequency domain
watermarking transforms the object into its frequency counterpart and
then embeds the watermark in the transform coefficients, distributing the
watermark over the entire frequency distribution of the object. The
frequency domain watermarking methods are relatively robust to noise,
image processing and compression compared with the spatial domain
methods [4, 6]. Extensive work has been done in both processing areas,
but watermarking techniques based on the transfer domain are more
popular than those of the spatial domain.
Types of Digital Watermarking
-
7/30/2019 watermarking algorithm
16/31
Watermarking Hardware Implementations:
Over the past decade, numerous watermarking algorithms have been
invented and their software is available, however recently, hardware
implementations are being presented in literature . Only a few hardware
schemes have been proposed. As proof of that, the following table of most of
the watermarking hardware implementations available in current literature.
Watermarking Hardware Implementations
Hardware implementations of watermarking can be implemented in
Application Specific Integrated Circuits (ASICs) or in FieldProgrammable Gate Arrays (FPGAs). As can be seen from the table
above, most of the current hardware implementations have been done
for ASIC designs. Recent advances in FPGA technology, such as
90nm process devices, higher gate densities, better interconnect
architectures, reduction in power consumption, multiple I/O formats
and embedded optimized logic, have allowed for applications that were
previously intended for ASICs to be implemented in FPGA devices,
with the added value of a lower FPGA cost when compared to an
ASIC. But for some reason that field has been understudied.
Watermarking implementation in hardware is a recent interest in the area. In
1999 there were still no works of video watermarking implementation in
hardware. Therefore, all of the works in this area are from the past 5
years, and nevertheless there are still very few works published.
Although the small amount of available works is a fact, the that there is a
-
7/30/2019 watermarking algorithm
17/31
strong motivation for hardware implementations because real-time
watermarking of video streams is too expensive for software. Audio and
Image watermarking are typically done in software because their low data
rates allow them to be processed in software.
As can be seen, watermarking implementations can be done in
software or in hardware. Although it might be faster to implement an
algorithm in software, there are a few compelling reasons for a move toward
hardware implementation. They add that in consumer electronic devices,
a hardware watermarking solution is often more economical because
adding the watermarking component takes up a small dedicated area of
silicon. In software, implementation requires the addition of a dedicated
processor such as a DSP core that occupies considerably more area,
consumes significantly more power, and may still not perform adequately
fast .
The Scientists presented a 0.18m CMOS technology implementation
of the Just Another Watermarking System (JAWS) embedder and
detector. The scientist selected this watermarking algorithm because it
is a well-known algorithm, because it works on raw video data
allowing the author to concentrate on the watermark process and not on the
compression issues, and because there was a previous implementation on
a Trimedia TM-1000 VLIW DSP done before, useful work to compare
their design . The JAWS processes uncompressed real-time video. The
authors claim that their work is the first step toward analyzing the
relationship between watermarking algorithmic features and
implementation cost for practical systems, and the first 0.18 micron
implementation published at that time. The implementation features a
pipelined architecture, and FFT and IFFT processing cores. The results show
watermarking of video streams at a rate of 30 frames/sec and 320 x 320
pixels/frame. The chip is capable of operating at 75 MHz and process a peak
pixel rate of over 3 MegaPixels/sec. The watermark is 4 bits/frame. The
power consumption for the embedder is 60 mW and for the detector is 100
mW.
-
7/30/2019 watermarking algorithm
18/31
Hardware implementation:
For the hardware implementation several design practices were taken into
account:
Word length: The word size used has an impact on the power
consumption of the hardware. Therefore, with a smaller word size,
less power dissipation we will have. The word size should be
limited to the smallest possible size in order to minimize power
consumption.
Frequency: CMOS-based circuits dissipate power only when a
transition from one logic state to another logic state occurs (although
recently this is not completely true due to higher sub-threshold
leakage currents).
Total power is determined by the sum of static power
dissipation and dynamic power dissipation. Static power is almost
zero (for this analysis) and dynamic power is , where is the
switching activity, or the percentage of time the circuit is switching,
C is the total load capacitance, V2 is the supply voltage squared,
and f is the operational frequency. In sequential circuits f is
regulated by the clock of the system, and the power consumption
of the system can be partially controlled by controlling f. A fastcircuit dissipates more power than a slow circuit [49]. In FPGA-
based implementations this is especially true due to the ability of
the device to allow the designer to determine the clock
frequency, and to design clock networks with various frequencies.
For low-power design a lower f is better. In areas of the system
where operating with a lower clock frequency is viable, the
option should be considered.
Switching: From the previous discussion about clock frequencyit can be inferred that if no switching occurs in the circuit, no
power is dissipated. Therefore switching should not be allowed
in a section of the system when that section is not in use. This
basically resumes to preventing the outputs of the circuits from
changing, by using chip enable controls.
-
7/30/2019 watermarking algorithm
19/31
Clock Gating: clock gating is a very popular low-power design
technique that is based on the previous discussion of the switching
principle. You can prevent circuit outputs from changing, but still
the clock network will be switching and dissipating power
although the circuits are not in use. Disconnecting the clockaltogether will prevent not only the circuits from switching and
dissipating power, but also clock network from dissipating power.
This is usually applied to whole sections of the system that are turned
off when not in use.
Area: Usually less area means less power, because fewer circuits
are needed to perform the work. More circuits mean more
switching, and therefore more power dissipation. Lowering area
typically means lowering performance in digital circuit design,
since fast-performing circuits generally require more logic.Arithmetic functions that utilize less logic are preferred in this
case. Area optimization techniques in the algorithmic description
should also be explored.
Fixed-point representation: Floating-point number representations
allow a larger dynamic range and precision, but fixed-point
representation is easier to implement, require less area, increases
speed, and decreases power consumption. Therefore a fixed-point
number representation should be used for a low-powerimplementation.
Arithmetic circuits: Arithmetic circuits should be employed that
reduce the number of operations performed in order to generate
a result, and that their switching activity is lower
Field programmable Gate Array (FPGA):
Field Programmable Gate Arrays, or FPGAs, are electronic devices that
contain a number of programmable logic, programmable interconnects,
and programmable I/O. This programmable logic can be used to
implement any logic based on logic gates such as ANDs, ORs, and
NOTs. Simple functions and complex functions, such as arithmetic
-
7/30/2019 watermarking algorithm
20/31
elements can be performed in an FPGA. They also provide memory
elements in the form of flip flops, or dedicated memory units. The high
density of programmable logic blocks found in FPGA and their relative
fast speeds have made them serious alternatives to Application Specific
Integrated Circuits (ASICs), microprocessors, and Digital Signal
Processors (DSPs). Also, their ability to perform any digital circuit,
their low time-to-market times, lower development costs, faster
debugging, and the re-programmability are additional benefits over
ASICs. An example of a simplified programmable logic block found in an
FPGA can be seen below.
FPGA programmable logic block
As can be seen from the picture above, the basic building block for
performing logic functions in FPGA are SRAMs, which are programmedas an n-input Look-Up Table (LUT).
The FPGA device used for this work is the Xilinx Spartan-3 X
C3S200ft256-4. It has 200,000 logic gates which translate to 4,320
programmable logic blocks, or cells. Also has 216k bits of dedicated
RAM, 12 18-bit-input dedicated multipliers, 4 Digital Clock Managers
with clock skew control, and frequency synthesis and shifting and 173
I/O pins with 18 single-ended programmable standards, and 8
differential-pair programmable standards. Xilinx products allow the LUTSRAMs to be also used as distributed RAM, or shift-registers. The device is
powered by a 1.32V internal voltage, a 3.00V auxiliary voltage and a
3.75V output driver supply voltage. The Xilinx Spartan-3 is marketed as
being the lowest-cost device per logic and I/O, and as a device that
can be used to develop high volume applications for a variety of
-
7/30/2019 watermarking algorithm
21/31
markets including broadband access, home networking,
display/projection and digital television equipment
Pseudo-Random Generator
The pseudo-random generator is implemented by using a Linear
Feedback Shift Register (LFSR). The LFSR provides the random
characteristics needed for the application [5, 12, 50]. The LFSR is 13-
bit wide register, thus producing a 2^13-1 or 8,191-bit long sequence (see
Figure 21). This modulates a total of ((2^13-1)/64), or 128, 8x8 blocks (one
block not completely watermarked). If we are using a 480x640 pixel
video frame, the available watermark bits are ((640x480)/64)/ 128, or 36,
bits. Therefore, according to these specifications, a 36 bit watermark word is
available for every frame.
These specifications can be varied easily making the watermark wordsmaller for increased watermark robustness. The LFSR used was a
Fibonacci LFSR with parallel load and chip enable generated using the
Xilinx Core Generator. The modulation of blocks is controlled by counters.
Eight binary values are generated consecutively, and these 8 values are
then used to select the 8-bit amplified value from a multiplexer. Since
the pseudo-random only generates 1s and 0s, we can think of the 0s as
being -1s. The amplification factor multiplied by 1 or -1 will result in the
amplification factor, or the negative of the amplification factor; therefore
we can take advantage of that and avoid a multiplication step. The
amplification factor, and its negative, is stored in a register and the
proper value is selected by the output of the pseudo-random generator.
These eight 8-bit amplified pseudo-random numbers are then passed on to
the DCT stage, to compute their Discrete Cosine Transform. This process
is repeated 8 times in order to generate a 8x8 block. The 13-bit
initialization value of the LFSR represents the key that only the watermark
decoder unit knows in order to retrieve the watermark from the video. If
each video camera in the surveillance system contains a different key then
we can identify the video stream from each camera when detecting the
watermark. The Pseudo-Random Generator outputs one number everyclock cycle. The 8 numbers are generated in 8 clock cycles; the 64 numbers
of the 8x8 block are generated in 64 clock cycles, but take 64 + 16
cycles to be read by the DCT core due to its timing limitations.
-
7/30/2019 watermarking algorithm
22/31
Watermarking Algorithm:
Our algorithm is invisible-robust algorithm which is implemented in
VLSI.In algorithm there are following notation :
: I original gray scale image, W binary or ternary watermark image,
-
7/30/2019 watermarking algorithm
23/31
I * watermarked image, (i, j) pixel location, E1 ,E2 watermark
embedding
functions, D watermark detection function, r neighborhood radius, IN
neighborhood image, K - digital watermark key, and a1 ,a2 scaling
constants.
The watermark insertion process consists of the following: First, the
watermark W
which is a ternary image having pixel values {0,1 or 2} is generated using
the digital key K. Then, watermark insertion is performed by altering the
pixels of original image using watermark embedding functions.
I*(i,j) = I(i,j) if W(i, j) = 0
I*(i,j) = E1 (I(i, j)), IN (i, j) ) if W(i, j) = 1
I*(i,j) = E1 ( I(i, j)), IN(i, j) ) if W(i, j) = 2
Where E1 and E2 are encoding function and these are defined as follows:
E1(I, IN ) = (1- a1)IN(i, j) + a1I(i, j)
E 2(I, IN ) = (1- a1) IN (i, j) + a2 I(i, j) The signs of
a1 and a2 are used for the detection function and their actual values
determine the watermark strength. The neighborhood image pixel gray value
IN is calculated as the average gray value of the neighboring pixels of the
original image for a neighborhood radius r. For example, for neighborhood
radius r =1.
IN(i, j) = [ {I(i+1,j)+I(i+1,j+1)}/2+ I(i, j + 1)]/2
The scaling (1 - a1 ) is used to scale IN to ensure that watermarked imagegray value I* never exceeds the maximum gray value for 8-bit image
representation correspond-ing to pure white pixel. The neighborhood radius
determines the upper bound of the watermarked pixels in an image.The first
step of detection process is the generation of watermark W using the wa-
termark key K. Next, the watermark is extracted from the test
-
7/30/2019 watermarking algorithm
24/31
(watermarked) image using the detection function given below, for a1 > 0
and a2 < 0.
W(i, j) = 1 ifI*(i, j) - IN(i, j) > 0
2 ifI*(i, j) - IN(i, j) < 0 ]
By comparing the original ternary watermark image W and the extracted
binary wa-termark image W* , the ownership can be established when the
detection ratio is largerthan a predefined threshold. The value of the
threshold determines the minimum ac-ceptable level of watermark
detection.
Architectural Design :
In this section, the architecture of the invisible-robust watermarking encoder
algorithm described in the previous section, is elaborated. We first provide
high level description of the enco der, followed by their architectural details.
Description of various terms:
1. Datapath and Controller:
-
7/30/2019 watermarking algorithm
25/31
The high-level view of the proposed chip is shown in Fig. 1. The encoder
includes the units, such as watermark generation, watermark insertion,
control, row and column address decoder, and registers. The generation unit
is used to produce the watermark, andinsertion unit is used to insert the
watermark into the host image as per the described algorithm. The control
unit controls the operationof the above two modules and the data .ow in
encoder. The address decoders are used to decode the memory address
where the image and watermark are stored. The registers are used for
buffering purpose. We assume that there are two external RAMs, one to
store the original image and other to serve as a storage space for watermark
data available. The watermarked image is written back to the RAM storing
the original image.
2 Watermark Generation Unit
The ternary watermark is generated by pseudorandom sequence generator.
The water-mark generation unit consists of linear feedback shift register
(LFSR). LFSR has a mul-titude of uses in digital system design and is a
very crucial unit in watermark security and detection. It is a sequential shift
register with combinational feedback logic around it that causes it to cycle
pseudo randomly through a sequence of binary values. Therefore, we have
studied the difficulties of a LFSR and have taken ap propriate measures to
ensure quality design [2123]. The LFSR consists of ip-ops (FFs) as
sequential elements with feedback loops. The feedback around a LFSR
comes from a selected set of points called taps in the FF chain and these
taps are fed back to FFs after either XORing or XNORing
-
7/30/2019 watermarking algorithm
26/31
Watermark Generation Unit: Linear Feedback Shift Register (LFSR).
The design aspects considered when modeling LFSRs are as follows [21
23]. XOR or XNOR Feed Back Gates: The feedback path may consistof
either all XORgates or all XNOR gates; LFSR will produce same number of
values with different sequence for a particular tap setting. One-to-Many or
Many-to-One Feedback Structure: Both one-to-many or many-to-one feed
back structures can be implemented using same number of gates. However, a
one-to-many feedback structure will have a shorter worst case delay. Pro
hibited or Lockup State: Special care should be placed on the design aspect
such that LFSR avoids the prohibited or lockup state. In the case of XORgates, the LFSR will not sequence through the binary value when all bits are
at logic zero. Similarly, for XNOR gates the LFSR will not sequence
through the binary values if all bits are at logic one. Thus, the LFSR should
bypass these initializations during power up.
Ensuring a Sequence of All 2n Values : If taps provided for a maximal
length se-quence are used, the LFSR configurations described so far will
sequence through (2n - 1) binary values. The feedback path can be modified
with extra circuitry to ensure that all 2n binary values are included in the
sequence. Fig. 2 shows the LFSR we designed adopting the abovediscussed facts. The 8-bit LFSR is modeled so as to use one-to-many
feedback structure and has been modified for a 2n looping sequence. It
calculates and holds the next value of the LFSR which is then assigned to
the output signal WM DATA after each clock edge. The NOR of all LFSR
bits minus the most significant bit that is LFSR REG (6:0) generates the
extra circuitry needed for all 2n sequence values.
-
7/30/2019 watermarking algorithm
27/31
3.3 Watermarking Insertion Unit
Fig. shows the architecture of the watermark insertion unit designed to
perform the watermarking insertion. The invisible-robust watermarking
involves adding or sub- tracting a constant times the pixel value to be
watermarked to or from a constant times the neighborhood function as
described in the watermark encoder function in the previous section. The
four data lines provide the pixels I(i, j), I(i +1,j), I (i, j +1),
-
7/30/2019 watermarking algorithm
28/31
-
7/30/2019 watermarking algorithm
29/31
-
7/30/2019 watermarking algorithm
30/31
-
7/30/2019 watermarking algorithm
31/31