Research On Digital Watermarking Algorithms In Still Image

With the rapid growth of internet technologies and extensive use of multimedia data, the enforcement of multimedia copyright protection becomes an important issue. Digital watermarking is viewed as an effective way for content authentication and copyright protection of multimedia data.This thesis focuses on images watermark algorithms. Firstly the research backgrounds, basic principles, representative algorithms and recent development of the digital watermarking technology are introduced. Then some techniques such as feature point extraction, quadtree decomposition, DFT, DCT and SVD, are adopted to the digital watermarking processing. Two algorithms for digital watermarking are proposed.One is an algorithm for localized watermarking based on quadtree decomposition. The main characteristics are as follows: (1) Combine transform domain algorithm with spatial domain algorithm, select some stable feature points in the image, use these points to quadtree decompose and mark independently different regions. In some suitable regions, insert watermark into the DFT domain. (2) It is a content-based watermark model. (3) In order to resist geometric distortions, a synchronization method based on image feature points, which is matched using relativity counting, is used to recover image size and aspect.The other is an algorithm of zero-watermark based on hybrid SVD and DCT. The main characteristics are as follows: (1) Sufficiently utilize the advantages of SVD and DCT, don't embed watermark in the original image, satisfy the imperceptibility and be robust to most attacks. (2) By adding external watermark, can not only hide the data, but also solve the high comparability issue of zero-watermark. (3) Exploiting the geometric moments of original image, estimate the geometric distortions parameters. So that the corrupted watermarked image can be corrected and the watermark can be detected correctly.Simulation results show that our algorithms are robust to many attacks. Besides, a watermark experiment system(WES) is designed and implemented to embed and detect watermark in the image.