| As an effective means to ensure the security of digital information, digital watermarking has received widespread attention from academia and industry community. Based on the application background and research status of digital watermarking, this paper focuses on a new multi-scale geometric analysis method-extended discrete Shearlet transform. The extended discrete Shearlet transform has directionality and anisotropy, it can not only well approach two-dimensional image, but also capture the directional characteristics of image more accurately, this paper proposes and completes two algorithm based on the horizontal cone of extended discrete Shearlet after testing and analyzing the decomposition characteristics of extended discrete Shearlet transform:(1)According to the basic principles and characteristics of extended discrete Shearlet transform, a digital watermarking algorithm based on human visual system in the horizontal cone of extended discrete Shearlet transform is proposed, which also uses the image contrast value theory, and the directional subgraph with maximum contrast value is selected for watermark embedding. Improve the visual masking model which is builded in Contourlet domain and then apply this model to horizontal cone domain of discrete Shearlet transform, and then a visual masking model which is suit for discrete Shearlet transform is obtained. Therefore, the biggest embedding strength can be obtained by calculating the Just Noticeable Difference value. Then embed the binary watermarking information into the selected low-frequency subband by the multiplicative principle. The watermark extracting is achieved by blind algorithm. Experiments show that the proposed algorithm has good robustness and imperceptibility.(2)This paper also proposes a watermarking algorithm combining extended discrete Shearlet transform and exponent moments (EMs), which utilizes the invariant of EMs and multi-direction features of extended discrete Shearlet transform. First, the extended discrete Shearlet transform is applied on host image. And then the EMs of the low-pass NSCT subband are computed. After computing, part of those EMs amplitude which are stable are selected. The digital watermark is embedded by quantizing the modulus of the selected EMs, and then watermarking algorithm based on local EMs in extended discrete Shearlet domain is proposed. Experimental results confirm that, our scheme can not only resist conventional attack, but also resist geometric distortions. |
