Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.570721
Title: A robust region-adaptive digital image watermarking system
Author: Song, Chunlin
Awarding Body: Liverpool John Moores University
Current Institution: Liverpool John Moores University
Date of Award: 2012
Availability of Full Text:
Access from EThOS:
Access from Institution:
Abstract:
Digital image watermarking techniques have drawn the attention of researchers and practitioners as a means of protecting copyright in digital images. The technique involves a subset of information-hiding technologies, which work by embedding information into a host image without perceptually altering the appearance of the host image. Despite progress in digital image watermarking technology, the main objectives of the majority of research in this area remain improvements in the imperceptibility and robustness of the watermark to attacks. Watermark attacks are often deliberately applied to a watermarked image in order to remove or destroy any watermark signals in the host data. The purpose of the attack is. aimed at disabling the copyright protection system offered by watermarking technology. Our research in the area of watermark attacks found a number of different types, which can be classified into a number of categories including removal attacks, geometry attacks, cryptographic attacks and protocol attacks. Our research also found that both pixel domain and transform domain watermarking techniques share similar levels of sensitivity to these attacks. The experiment conducted to analyse the effects of different attacks on watermarked data provided us with the conclusion that each attack affects the high and low frequency part of the watermarked image spectrum differently. Furthermore, the findings also showed that the effects of an attack can be alleviated by using a watermark image with a similar frequency spectrum to that of the host image. The results of this experiment led us to a hypothesis that would be proven by applying a watermark embedding technique which takes into account all of the above phenomena. We call this technique 'region-adaptive watermarking'. Region-adaptive watermarking is a novel embedding technique where the watermark data is embedded in different regions of the host image. The embedding algorithms use discrete wavelet transforms and a combination of discrete wavelet transforms and singular value decomposition, respectively. This technique is derived from the earlier hypothesis that the robustness of a watermarking process can be improved by using watermark data in the frequency spectrum that are not too dissimilar to that of the host data. To facilitate this, the technique utilises dual watermarking technologies and embeds parts of the watermark images into selected regions of the host image. Our experiment shows that our technique improves the robustness of the watermark data to image processing and geometric attacks, thus validating the earlier hypothesis. In addition to improving the robustness of the watermark to attacks, we can also show a novel use for the region-adaptive watermarking technique as a means of detecting whether certain types of attack have occurred. This is a unique feature of our watermarking algorithm, which separates it from other state-of-the-art techniques. The watermark detection process uses coefficients derived from the region-adaptive watermarking algorithm in a linear classifier. The experiment conducted to validate this feature shows that, on average, 94.5% of all watermark attacks can be correctly detected and identified.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.570721  DOI: Not available
Keywords: QA75 Electronic computers. Computer science ; QA76 Computer software
Share: