Wavelet-based Digital Watermark

The development of multimedia storage and transmission technology, especially the popularization of the network technology, provides much convenience to the spread of digital information and pushes the development of digital media application, but it also brings us some side effects and causes new problems. People can find the image and audio files easily from the Internet and they can edit, modify, duplicate and spread these files, thus invading the author's copyright. Therefore, it is an urgent problem to be solved about how to protect the copyright and economic benefits.Under such circumstance, digital watermarking technique has been investigated. Digital watermarking technique is the frontier field of academia in recent years. And it certificates the ownership of copyrights or tracks tort by embedding some secret information in the protected digital objects, such as images and audios.In this thesis, we mainly discuss the digital image and audio watermarking algorithms in wavelet domain. Wavelet transform has a multi-resolution analytic character. And it can fetch information and analyze functions or signals in a multi-resolution way through kinds of operations. Moreover, both of its time-window and frequency-window can be changed. It is able to express local characters of information in time-frequency domain.In the first place, the thesis introduces a digital image watermarking method combining Discrete Wavelet Transform (DWT) with Discrete Cosine Transform (DCT). Firstly, the original digital watermarking image is processed by toral automorphisms and Zigzag before embedded, thus decreasing or eliminating correlation among pixels. Secondly, the embedding algorithm based on DWT and DCT has the advantages of both DWT and DCT. When it comes to the process of watermarking extraction, the thesis provides an improved extraction algorithm to reduce the damage to images' quality caused by the storage format and computational accuracy.In the second place, a new un-embedding digital watermarking algorithm is proposed in the thesis. During the process of "embedding", the original carrier image is not been modified. And the thesis proposes a new method to choose the operating bits dynamically and combines this method with the improved strategy of mean comparison. In this way, it is certificated by experiments that such un-embedding method is very robust.At last, the thesis proposes an audio watermarking algorithm with adaptive quantization step size and embedding position. In the algorithm, the original audio is divided into several sections. Then, different quantization step size is used based on different audio sections, and watermarking is embedded in the optimal position adaptively. This algorithm is not only transparent but also robust.