LAUSR

AbstractRobustness has played extremely important role in the multiple applications of digital watermarking technology. The reason lies in its influence on the watermarking system’s practicability. As one of the most difficult kinds of digital signal processing for a digital watermark to survive, geometric distortions have become a central problem in digital image watermarking research. Therefore, resigning a greatly robust digital image watermarking approach which can withstand geometric distortions is still a quite challenging work. On account of Bessel K-form (BKF) probability density function (PDF), this paper proposes an optimal synchronization correction-based digital image watermarking method which can resist geometric distortions. This approach consists of watermark embedding, synchronization correction and watermark extraction. Using the quantization index modulation (QIM), the watermark is inserted into the original host images in the nonsubsampled shearlet transform (NSST) domain by adjusting the selected blocks’ low-frequency NSST coefficients. The BKF PDF describes the NSST coefficients’ statistical features. Besides, the BKF statistical model parameters can be used in constructing a compact image characteristic space. Utilizing the compact image feature, the least squares support vector regression (LS-SVR) synchronization correction is performed to estimate the geometric distortions parameters. After LS-SVR synchronization correction, the inverse QIM is performed to recover blindly the inserted watermark. Simulation results demonstrate that the presented digital image watermarking method not only has good imperceptibility performance, but also can well resist challenging common signal processing operations as well as geometric distortions, which can be superior to the most advanced approaches.