Watermarking Techniques For Digital Images

As a useful tool for the intellectual copyright protection, watermarking technique has gained more and more concerns in many ways since it appeared in 1993. In this paper, we summarize the general theories of the digital watermark, analysis the key points about watermarking technique, and propose two more practical watermark algorithms based on former studies: one based on image gradients and the other based on wavelet transfonn. At the same time, we implement some experiments to test invisibility and robustness of the watermarks produced from the above two algorithms. The algorithm based on image gradients makes good use of some content of images--the edge--to embed and extract the watermark. Since the HVS has especial responses on edges, the watermark embedded on edges has good robustness to general image processing and other attacks. In order to have more robustness to the geometrical attacks such as cropping and not to depend on the original image excessively in spatial domain, the watermark is permuted twice before embedded. Moreover, the extracted watermark is post processed by the crux middle filter. The other algorithm based on wavelet transform uses the idea of the Embedded Zero-tree Wavelet Code for reference. The watermark is embedded to the corresponding different resolution areas of different bit planes of the wavelet image. The embedded capacity factor is determined by the area and bit plane where the selected wavelet coefficient is. By the layers based on bit planes, this algorithm complement the embedding of two different types of watermarks: the two-valued sign image and the pseudorandom sequence according to N(O, 1). The extraction of watermarks is implemented by the bit plane layer. Especially, to the extraction of the pseudorandom sequence, the extracted watermark is tested by the sequence expansion along the bit plane. Because the test is implemented by the similarit\ between the extracted sequence and the original sequence, the test can stop immediately when the similarity value reached the initialized threshold. Therefore. this extraction method can ensure the veracity of the extracted watermark and improve the efficiency of extracting. In the same way as the algorithm based O[1 image gradients, to the two-valued sign watermark, the retrieved watermark is post processed by the crux median filter in order to improve the quality of the retrieved watermark. At the same time, it is easier to combine this algorithm with the wavelet-based compression algorithm so as to implement the embedment of the watermark in the process of image compression coding. The study in this paper shows that there are many problems that need to resolve in the development of digital watermarks. The watermark technique is a promised method in the intellectual copyright protection during the multimedia trading by Internet. Zhang Sujuan (Pattern Recognition and Image Processing) Directed by Prof. Zhang Hongbin...