基于LWT-QR分解和Laplacian Pyramid的水印算法

目的为了提高数字水印的鲁棒性和不可见性,提出一种基于Laplacian Pyramid和LWT-QR分解的水印算法。方法首先对宿主图像进行2层Laplacian Pyramid分解,取其第2层Laplacian残差图像进行一层LWT分解,取其低频子带进行大小为4×4的无重叠分块处理。然后,基于提升小波系数的相关属性,再对每个选中的低频子块进行QR分解,取分解后R矩阵的第1行为目标进行水印的嵌入,同时对水印进行Arnold置乱,置乱后的水印图像嵌入到R矩阵的第1行元素中。结果嵌入水印后图像的PSNR能够达到45 d B,而且该方案对常见的信号处理攻击有较好的鲁棒性,NC均值在0.9以上。结论理论分析和大量的实验数据表明,该方案能够很好地改善图像操作过程中的鲁棒性和不可见性。

Watermarking Algorithm Based on the LWT-QR Decomposition and Laplacian Pyramid

The work aims to improve the robustness and invisibility of digital watermarking and put forward a kind of watermarking algorithm based on the Laplacian Pyramid and LWT-QR decomposition. First, the two layers'' Laplacian Pyramid decomposition of host image was conducted. The second layer of Laplacian residual image was taken for a layer of LWT decomposition, and the low frequency subband size was subject to the 4×4 non-overlapping block processing. Then, based on the related properties used to improve the wavelet coefficient, the QR decomposition of each selected line frequency sub-block was done. The watermark was embedded with the first line of the decomposed R matrix as the target. Meanwhile, Arnold scrambling of the watermark was conducted. The scrambled watermark image was embedded into the first-line elements of R matrix. The PSNR of the watermarked images could reach 45 dB, and such scheme had better robustness against the common signal processing attacks with the average NC values all above 0.9. Theoretical analysis and a lot of experimental data show that the scheme can well improve the robustness and invisibility in the operation process of the image.