Previous research [Cox97b] indicates that significant portions of the host image, e.g. the low-frequency components, have to be modified in order to embed the information in a reliable and robust way. This led to the development of watermarking schemes embedding in the frequency domain. Nevertheless, robust watermarking in the spatial domain can be achieved [Bruyndonckx95a] at the cost of explicitly modeling the local image characteristics. These characteristics can be more easily obtained in a transform domain, however.
Many image transforms have been considered, most prominent among them is the discrete cosine transform (DCT) which has also been favored in the early image and video coding standards. Hence, there is a large number of watermarking algorithms that use either a block-based [Koch95a,Bors98] or global DCT [Cox97b,Barni98d]. Other transforms that have been proposed for watermarking purposes include the discrete Fourier transform (DFT) [Ramkumar99m], the Fourier-Mellin transform [Ruanaidh98b] and the fractal transform [Puate96a,Dugelay99a]. In this work, we focus on the wavelet domain for the reasons given below.
With the standardization process of JPEG2000 and the shift from DCT- to wavelet-based image compression methods, watermarking schemes operating in the wavelet transform domain have become even more interesting. New requirements such as progressive and low bit-rate transmission, resolution and quality scalability, error resilience and region-of-interest (ROI) coding have demanded more efficient and versatile image coding [Charrier99a]. These requirements have been met by the wavelet-based "Embedded Block Coding with Optimized Truncation" (EBCOT) system [Taubman99a], which was accepted with minor modifications as the upcoming JPEG2000 image coding standard [jpeg2000fcd]. The wavelet transform [Daubechies92a] has a number of advantages [Xia98a,Lumini00a] over other transforms such as the DCT that can be exploited for both, image compression and watermarking applications. Therefore, we think it is imperative to consider the wavelet transform domain for watermarking applications.
Chapter 3 starts by first characterizing the most important and distinguishing features of previously proposed wavelet-based watermarking schemes. We organize the overwhelming amount of algorithms proposed in the literature in two main categories: additive and quantize-and-replace strategy embedding techniques. Further on, each approach is discussed in detail, building on the experience that was gained from implementing some of the watermarking schemes.
Our own contributions are presented in chapter 4. First, we propose using wavelet filter parametrization as a means to improve the security of many previously proposed algorithms and demonstrate that our concept of secret key-dependent wavelet filters can be easily employed as a security framework. Motivated by the increasing importance of the forthcoming JPEG2000 image standard, we present a novel watermarking schemes that is compatible with the independent code-block processing approach of the new image coding standard. Two application scenarios, copyright protection and image authentication, are considered and we demonstrate that our quantization-based embedding technique can successfully encode and decode a binary watermark on-the-fly in the JPEG2000 coding process.
In chapter 5, a classification of attacks on watermarks is given which will be used to discuss the robustness results in chapter 6.
Slides available in Acrobat (.PDF, zipped) und Postscript (.ps.gz, gzipped) format.