Statistical image watermark decoder by modeling local RDWT difference domain singular values with bivariate weighted Weibull distribution

For any watermarking scheme, how to simultaneously improve robustness, invisibility and watermarking capacity has always been a problem that the scientific research community is committed to solving. Statistical modeling technology is an effective means to solve this problem. In this paper, a new method of digital image watermarking technology is proposed for the first time by using Bivariate Weighted Weibull distribution to model Redundant Discrete Wavelet Transform (RDWT) difference domain local Singular Value (SVD) coefficients. First, according to the decomposition characteristics of RDWT, the difference between the first scale and the second scale subband is calculated, and local SVD is performed on the difference subband to obtain the modeling object. In the experiment, the embedding carrier is nonlinear. Although the linear embedding method is still used in the embedding process, we use the entropy value to modify the different number of coefficients in different coefficient blocks to embed watermark bits, so it meets the requirements of image nonlinear characteristics. Then, the Bivariate Weighted Weibull distribution is used for modeling and constructing the decoder, which can effectively take advantage of the strong correlation between the subband directions of the difference. Finally, through a large number of simulation experiments, the performance of the proposed algorithm is comprehensively evaluated. The verification results confirm that the proposed method can still obtain excellent anti-attack effect and invisibility when the watermark capacity is large.