19-07-2013, 04:55 PM
Digital Watermarking
Digital Watermarking.pdf (Size: 836.89 KB / Downloads: 325)
Introduction
Digital watermarking is the act of hiding a message related to a digital signal (i.e. an
image, song, video) within the signal itself. It is a concept closely related to
steganography, in that they both hide a message inside a digital signal. However, what
separates them is their goal. Watermarking tries to hide a message related to the actual
content of the digital signal, while in steganography the digital signal has no relation
to the message, and it is merely used as a cover to hide its existence.
Watermarking has been around for several centuries, in the form of watermarks found
initially in plain paper and subsequently in paper bills. However, the field of digital
watermarking was only developed during the last 15 years and it is now being used
for many different applications.
In the following sections I will present some of the most important applications of
digital watermarking, explain some key properties that are desirable in a
watermarking system, and give an overview of the most common models of
watermarking as presented in the book by Ingemar Cox, Matthew Miller, Jeffrey
Bloom, Jessica Friedrich and Ton Kalker [1]. These basic models will be further
illustrated by the use of example watermarking systems that were developed in
Matlab. All images used in this essay, except those used to present the results of the
example watermarking systems are taken from this book [1].
Watermarking applications
The increasing amount of research on watermarking over the past decade has been
largely driven by its important applications in digital copyrights management and
protection.
One of the first applications for watermarking was broadcast monitoring. It is often
crucially important that we are able to track when a specific video is being broadcast
by a TV station. This is important to advertising agencies that want to ensure that their
commercials are getting the air time they paid for. Watermarking can be used for this
purpose. Information used to identify individual videos could be embedded in the
videos themselves using watermarking, making broadcast monitoring easier.
Watermarking properties
Every watermarking system has some very important desirable properties. Some of
these properties are often conflicting and we are often forced to accept some trade-
offs between these properties depending on the application of the watermarking
system.
The first and perhaps most important property is effectiveness. This is the probability
that the message in a watermarked image will be correctly detected. We ideally need
this probability to be 1.
Another important property is the image fidelity. Watermarking is a process that alters
an original image to add a message to it, therefore it inevitably affects the image’s
quality. We want to keep this degradation of the image’s quality to a minimum, so no
obvious difference in the image’s fidelity can be noticed.
The third property is the payload size. Every watermarked work is used to carry a
message. The size of this message is often important as many systems require a
relatively big payload to be embedded in a cover work. There are of course
applications that only need a single bit to be embedded.
The false positive rate is also very important to watermarking systems. This is the
number of digital works that are identified to have a watermark embedded when in
fact they have no watermark embedded. This should be kept very low for
watermarking systems.
Communication-based models
Communication-based models describe watermarking in a way very similar to the
traditional models of communication systems. Watermarking is in fact a process of
communicating a message from the watermarking embedder to the watermarking
receiver. Therefore, it makes sense to use the models of secure communication to
model this process.
Geometric models
It is often useful to think of watermarking in geometric terms. In this type of model,
images, watermarked and unwatermarked, can be viewed as high-dimensional
vectors, in what is called the media space. This is also a high-dimensional space that
contains all possible images of all dimensions. For example a 512 X 512 image would
be described as a 262144 elements vector in a 262144-dimensional space.
Geometric models can be very useful to better visualize the watermarking process
using a number of regions based on the desirable properties of watermarking. One of
these regions is the embedding region, which is the region that contains all the
possible images resulting from the embedding of a message inside an unwatermarked
image using some watermark embedding algorithm. Another very important region is
the detection region, which is the region containing all the possible images from
which a watermark can be successfully extracted using a watermark detection
algorithm. Lastly, the region of acceptable fidelity contains all the possible images
resulting from the embedding of a message into an unwatermarked image, which
essentially look identical to the original image.
Conclusion
Watermarking is a very active research field with a lot of applications. Although it is a
relatively new field, it has produced important algorithms for hiding messages into
digital signals. These can be described by many different models. Two broad
categories for these models were described in this essay. These are communication-
based models and geometric models. Communication-based models can be further
divided into those which use side-information and those that don’t. One example
system was used to illustrate non-side-information models, and two example systems
were used to illustrate side-information models. Each of these systems has its
advantages and disadvantages, and each one trades some important watermarking
property for another. The choice of which to use relies on the underlying application’s
requirements.
Digital Watermarking.pdf (Size: 836.89 KB / Downloads: 325)
Introduction
Digital watermarking is the act of hiding a message related to a digital signal (i.e. an
image, song, video) within the signal itself. It is a concept closely related to
steganography, in that they both hide a message inside a digital signal. However, what
separates them is their goal. Watermarking tries to hide a message related to the actual
content of the digital signal, while in steganography the digital signal has no relation
to the message, and it is merely used as a cover to hide its existence.
Watermarking has been around for several centuries, in the form of watermarks found
initially in plain paper and subsequently in paper bills. However, the field of digital
watermarking was only developed during the last 15 years and it is now being used
for many different applications.
In the following sections I will present some of the most important applications of
digital watermarking, explain some key properties that are desirable in a
watermarking system, and give an overview of the most common models of
watermarking as presented in the book by Ingemar Cox, Matthew Miller, Jeffrey
Bloom, Jessica Friedrich and Ton Kalker [1]. These basic models will be further
illustrated by the use of example watermarking systems that were developed in
Matlab. All images used in this essay, except those used to present the results of the
example watermarking systems are taken from this book [1].
Watermarking applications
The increasing amount of research on watermarking over the past decade has been
largely driven by its important applications in digital copyrights management and
protection.
One of the first applications for watermarking was broadcast monitoring. It is often
crucially important that we are able to track when a specific video is being broadcast
by a TV station. This is important to advertising agencies that want to ensure that their
commercials are getting the air time they paid for. Watermarking can be used for this
purpose. Information used to identify individual videos could be embedded in the
videos themselves using watermarking, making broadcast monitoring easier.
Watermarking properties
Every watermarking system has some very important desirable properties. Some of
these properties are often conflicting and we are often forced to accept some trade-
offs between these properties depending on the application of the watermarking
system.
The first and perhaps most important property is effectiveness. This is the probability
that the message in a watermarked image will be correctly detected. We ideally need
this probability to be 1.
Another important property is the image fidelity. Watermarking is a process that alters
an original image to add a message to it, therefore it inevitably affects the image’s
quality. We want to keep this degradation of the image’s quality to a minimum, so no
obvious difference in the image’s fidelity can be noticed.
The third property is the payload size. Every watermarked work is used to carry a
message. The size of this message is often important as many systems require a
relatively big payload to be embedded in a cover work. There are of course
applications that only need a single bit to be embedded.
The false positive rate is also very important to watermarking systems. This is the
number of digital works that are identified to have a watermark embedded when in
fact they have no watermark embedded. This should be kept very low for
watermarking systems.
Communication-based models
Communication-based models describe watermarking in a way very similar to the
traditional models of communication systems. Watermarking is in fact a process of
communicating a message from the watermarking embedder to the watermarking
receiver. Therefore, it makes sense to use the models of secure communication to
model this process.
Geometric models
It is often useful to think of watermarking in geometric terms. In this type of model,
images, watermarked and unwatermarked, can be viewed as high-dimensional
vectors, in what is called the media space. This is also a high-dimensional space that
contains all possible images of all dimensions. For example a 512 X 512 image would
be described as a 262144 elements vector in a 262144-dimensional space.
Geometric models can be very useful to better visualize the watermarking process
using a number of regions based on the desirable properties of watermarking. One of
these regions is the embedding region, which is the region that contains all the
possible images resulting from the embedding of a message inside an unwatermarked
image using some watermark embedding algorithm. Another very important region is
the detection region, which is the region containing all the possible images from
which a watermark can be successfully extracted using a watermark detection
algorithm. Lastly, the region of acceptable fidelity contains all the possible images
resulting from the embedding of a message into an unwatermarked image, which
essentially look identical to the original image.
Conclusion
Watermarking is a very active research field with a lot of applications. Although it is a
relatively new field, it has produced important algorithms for hiding messages into
digital signals. These can be described by many different models. Two broad
categories for these models were described in this essay. These are communication-
based models and geometric models. Communication-based models can be further
divided into those which use side-information and those that don’t. One example
system was used to illustrate non-side-information models, and two example systems
were used to illustrate side-information models. Each of these systems has its
advantages and disadvantages, and each one trades some important watermarking
property for another. The choice of which to use relies on the underlying application’s
requirements.