Research On Digital Watermarking Scheme Using SVD In Non-subsampled Shearlet Transform Domain

At present,the multi-scale analysis technology is widely used in the image processing field,and the combination with digital watermarking technology is becoming a hotspot.In recent years,a novel Shearlet transform based on multi-scale analysis is proposed.The Shearlet transform has the advantages of anisotropy,multi-resolution and locality.Besides,it can not only deal with the singularity information of multi-dimensional functions,but also obtain the sparse representation and nonlinear approximation of images.It can obtain better invisibility and robustness when it is applied to the digital watermarking.This thesis proposes two kinds of digital watermarking algorithms after analyzing the characteristics of the non-subsampled Shearlet transform.The main contents of this thesis are as follows:(1)Research on non-subsampled Shearlet transform.The realization process of non-subsamp-led Shearlet transform and its advantages are analyzed,which provide the theoretical basis for the combination with the digital watermarking.(2)Research on watermarking embedding position and method based on non-subsampled Sh-earlet transform.Considering that the Human Visual System has different sensitivity about the smooth and non-smooth parts of the image,this thesis proposes a novel block-based digital watermarking scheme in non-subsampled Shearlet transform domain.The original image is divided into non-overlapping blocks with the same size,and the target blocks are selected according to a certain principle.The non-subsampled Shearlet transform is applied to the selected parts of the image to obtain the sub-band images with the different resolutions and different directions which provide a lot of positions for embedding watermarking information.The appropriate sub-band images are chosen to embed the watermarking information by utilizing the stability of singular value decomposition.Finally,the experiments show that the proposed algorithm has good robustness of resisting various attacks such as cropping,filtering and JPEG compression attacks.(3)A digital watermarking algorithm based on Harris-Laplace and genetic algorithm in non-subsampled Shearlet transform domain is proposed.Harris-Laplace is used to detect the feature points which stands for the characteristics of the image and calculate the feature scales.The non-subsampled Shearlet transform is applied to the stable and non-overlapping geometrically invariant regions.In the embedding algorithm,this process is carried out by quantization the sub-band coefficients.We can obtain the optimal embedding strength under the eight different kinds of attacks by utilizing the genetic optimization algorithm,which ensures this thesis acquires good Peak Signal Noise Ratio and the Normalized Correlation.Finally,the invisibility and robustness can achieve good balance.The results show that the algorithm achieves good performance and obtains good invisibility and robustness at the same time.