Research On Digital Image Watermarking Technique

In the last decade, the fast development of computer science, communication technology and multimedia application technology makes variety of multimedia widely used, such as digital audio, digital image, digital video, E-book, etc. These digital products bring us a lot of convenience and pleasure. However, they can be replicated, edited easily and spread openly through Internet or by other means, which causes a serious problem that it is difficult even impossible to give proof of copyright once pirated, as a result, the copyright is unprotected. To solve these kinds of problems, some people believe that it is useful to hide some information called watermark into the digital products for the purpose of copyright protection.As the variety of watermark and its wide use, this thesis will focus on the three following aspects: image scrambling, color image watermark, and video watermark.Firstly, the idea of code space associated with IFS is used for image scrambling. In this strategy, the same image is coded by two different rules, then the pixels' position of image is changed by a transform between two different pixel (or point) address code spaces. Experimental result shows that this scheme is useful for image scrambling. It is characterized by the satisfactory scrambling result just after one transformation.Secondly, at present, most of watermarking algorithms are based on DFT, DCT and DWT. Watermark information is embedded into some frequency band after decomposition. In transform domain, most of energy is located at the low frequency area, watermark embedded here is robust, but watermark data and its embedding strength are quite limited. On the other hand, rich texture and edge information are located at high frequency area, it allows hiding more watermark information, however it is vulnerable to some manipulation (data compression, especially). According to the image similarity (similar images have similar distribution in transform domain.) and the linearity of the above three transforms, a method is put forward based on the similarity of two image blocks between watermark image and host image. Watermark image blocks are embedded into the host image blocks according to the best similarity coefficient. Because of the similarity between them, watermark blocks are embedded well invisibly and show great robust. As demonstrated in experiment, digital watermark is robust against cutting, noising, lossy JPEG compression, blurring, sharpening, rotation, and resizing. At the same time, it is believed that this...