摘要

Protection against geometric distortions and common image processing operations with blind detection becomes a much challenging task in image watermarking. To achieve this, in this paper we propose a content-based watermarking scheme that combines the invariant feature extraction with watermark embedding by using Tchebichef moments. Harris-Laplace detector is first adopted to extract feature points, and then non-overlapped disks centered at feature points are generated. These disks are invariant to scaling and translation distortions. For each disk, orientation alignment is then performed to achieve rotation invariant. Finally, the watermark is embedded in magnitudes of Tchebichef moments of each disk via dither modulation to realize the robustness to common image processing operations and the blind detection. Thorough simulation results obtained by using the standard benchmark, Stirmark, demonstrate that the proposed method is robust against various geometric distortions as well as common image processing operations and outperforms representative image watermarking schemes.