Robust Watermarking Methods Based On Statistical Features

With the rapid growth of the Internet and computer technologies, the copyright protection of digital multimedia has received increasing interest. Digital watermarking, as an important and promising technology, has been proposed to tackle this problem. Among the existing watermarking methods, however, there are few methods consider-ing the robustness against complex geometric transformation including rotation, scaling and cropping except common image processing, e.g., JPEG compression, low filtering. For this purpose, this paper first proposes a robust watermarking method by utilizing the statistical features of images, i.e., histogram, for watermark embedding, which outper-forms the classical methods in terms of robustness and imperceptibility. To further ex-pand the application of the aforementioned method, we combine it with quantitative embedding model in frequency domain to design another robust method for color im-ages. Since different embedding models focus on different kinds of attacks, the second method significantly improve the robustness.The method based on histogram has been demonstrated to be robust against com-plex geometric attacks, which lies in the following properties: first, the watermark em-bedding based histogram is approximately invariant to complex geometric transforma-tion; secondly, global and local search strategies are applied to detect the watermarks, avoiding the synchronization errors and improving the precision of detection.In view of the widespread applications of color images, we carefully design a ro-bust watermarking method for color images, which not only utilizes the merits of dif-ferent embedding models in spatial and frequency domains, but also considers the per-ceptual characteristics of human visual system (HVS) to different color components. Based on this, the proposed method repeatedly embeds the same watermarks into the different components of a color image by using different embedding models, leading to better robustness and imperceptibility.The experimental results conducted on standard images demonstrate that the effec-tiveness and superiority of the proposed methods. In summary, they achieve stronger robustness against common image processing and complex geometric attacks while preserving good imperceptibility.