الفهرس يوجد فقط 14 صفحة متاحة للعرض العام
 |  | from | 146 |  |  |
| | | from | 146 | | |
المستخلص
Due to the powerful computing capabilities of modern
computers and the increasing popularity of the Internet, digital
multimedia products are used and distributed much easier and
faster. However, illegal copying and manipulation of such
media can cause considerable financial loss. As a result,
copyright protection of digital media has recently become an
active research area.
Recently, digital watermarking has been extensively
studied for copyright protection of digital images. Basically,
digital watermarking is the process of embedding hidden
copyright information, the watermark, into digital data by
making small modifications on these data without destroying
its value.
There are two methods of performing image
watermarking, one in spatial domain and the other in transform
domain. In the spatial domain we can simply insert watermarks
into a host image by changing the gray levels of some pixels in
the host image. In transform domain, we can insert watermarks
into coefficients of a transformed image, for example, using
Discrete Fourier Transform (DFT) domain, Discrete Cosine
Transform (DCT) domain or Discrete Wavelet Transform
(DWT) domain.
The problem of choosing the best domain for embedding
watermarks in digital images is very important. The main
objective of this work is to make a study on digital image
watermarking techniques in order to specify the best domain
for embedding watermarks in digital images with respect to the
different applications of digital image watermarking.
In this thesis, seven image watermarking techniques are
implemented and evaluated. The evaluated techniques are
selected to represent the different approaches to embedding
data in spatial domain, DFT domain, DCT domain, and DWT
domain. The algorithms are also chosen to represent a range of
computational complexities and implementation structures. The
performance of the selected algorithms is evaluated with
respect to perceptual quality, execution time, and robustness in
face of nine different attacks. The chosen attacks are
commonly used.
The simulation results and performance analysis show that
the DWT domain is the most promising domain for embedding
watermarks in digital images. The results show also that the
two requirements, robustness and imperceptibility, are contrary
to each other. Generally speaking, in the transform-domain
watermarking techniques, embedding the watermark into the
higher frequency coefficients is not robust, although the
watermarked image quality is assured. In contrast, embedding
the watermark into the lower frequency coefficients is more
robust against common attacks such as low pass filtering and
lossy compression but it will cause the resulting watermarked
image quality greatly degrades to compare with the original
image.
The second contribution of this thesis is that, a wavelet-
based watermarking scheme combined with genetic algorithms
(GA) is introduced. The objective of the proposed scheme is to
improve the performance, not only in watermarked image
quality, but also in robustness against a wide range of attacks
using genetic training techniques. A fitness function based on
the two conflicting requirements is defined and GA is used to
optimize it. The simulation results and performance analysis
show that the application of genetic algorithms in the
watermarking problem is promising.