Hai, Tao creator text my Kuantan, Pahang UMP 2012 monographic eng

print

xvii, 155 p. : ill. (some col.) ; 30 cm. + 1 CD-ROM Digital watermarking techniques have been explored considerably since its first appearance in the 1990s. The achieved tradeoffs from these techniques between imperceptibility and robustness are controversial. To solve this problem, this study proposes the application of artificial intelligent techniques into digital watermarking by using discrete wavelet transform (DWT) and singular value decomposition (SVD). To protect the copyright information of digital images, the original image is decomposed according to two-dimensional discrete wavelet transform. Subsequently the preprocessed watermark with an affined scrambling transform is embedded into the vertical subband (HLm) coefficients in wavelet domain without compromising the quality of the image. The scaling factors are trained with the assistance of Particle Swarm Optimization (PSO). A new algorithmic framework is used to forecast feasibility of hypothesized watermarked images. In addition, the novelty is to associate the Hybrid Particle Swarm Optimization (HPSO), instead of a single optimization, as a model with SVD. To embed and extract the watermark, the singular values of the blocked host image are modified according to the watermark and scaling factors. A series of training patterns are constructed by employing between two images. Moreover, the work takes accomplishing maximum robustness and transparency into consideration. HPSO method is used to estimate the multiple parameters involved in the model. Unfortunately, watermark resistance to geometric attacks is the most challenge work in traditional digital image watermarking techniques which causes incorrect watermark detection and extraction. Recently, the strategy of researchers has introduced image watermarking techniques using the invariant transforms for their rotation and scale invariant properties. However, it suffers from local transformations which make watermarks difficult to recover. This thesis will introduce a set of content based image watermarking schemes which can resist both local geometric attacks and traditional signal processingattacks simultaneously. These schemes follow a uniform framework, which is based on the detection of feature points which are commonly invariant to Rotation, Scaling and Translation (RST), therefore they naturally accommodate the framework of geometrically robust image watermarking. As a result, it will first introduce the theories about the feature extraction and the basic principles on how feature points can act as locating resynchronization between watermark insertion and extraction discussed in detail. Subsequently, it will present several content-based watermark embedding and extraction methods which can be directly implemented based on the synchronization scheme. Further detailed watermarking schemes which combine feature regions extraction with counter propagation neural network-based watermarks synapses memorization are then presented. The performance of watermarking schemes based on framework of feature point shows the following advantages: (a) Good imperceptibility. It is obvious that the watermarking schemes show a little influence on watermark invisibility; (b) Good robustness. The proposed scheme is not only robust against common image processing operations as sharpening, noise adding, and JPEG compression etc, but also robust against the desynchronization attacks such as rotation, translation, scaling, row or column removal, cropping, and local random bend etc. specialized Hai Tao Thesis (Doctor of Philosophy (Computer Science)) -- Universiti Malaysia Pahang - 2012 Bibliography : p. 137-147 Digital watermarking Watermarking Data protection Computer security THE0001929(Local) http://ecollib.ump.edu.my/3678/ http://ecollib.ump.edu.my/3678/ UMP 120911 20251114204520.0 vtls000063638