基于脊波变换的量化水印算法

Ridgelet transform based quantitative watermarking algorithm

利用比小波变换域更适合表示图像边缘轮廓特征的脊波变换域设计一种新的图像水印算法.将图像分块脊波变换,在脊波系数最大能量方向选择细节系数,设计合理的量化水印方法.采用奇偶量化方法嵌入水印,利用归属量化区间法提取水印.在较强的攻击下,该方法能够很好地恢复原始水印,采用该方法嵌入水印不易导致水印化图像失真.攻击实验结果表明,与小波域水印算法相比,该算法不仅可以较好地抵抗各种常见攻击如压缩、加噪、滤波和剪切等,还能较好地抵抗几何旋转攻击.

A novel image watermarking algorithm was designed by taking the advantage that the ridgelet transform domain is a more suitable representation of image＇s outline edge features than the wavelet transform domain. First, images were divided block by block, and each block was treated by finite ridgelet transform. Second, the direction of the largest energy of ridgelet transform coefficients was found for each block. At last, a reasonable quantitative watermarking method was used to the coefficients for embedding watermark in the largest energy direction. Odd-even quantitative method was used to embed watermark, and quantitative section attributive method was used to extract the original watermark. The watermark extraction method can efficiently restore the original watermark even if suffered terrible attack. The embed watermark in the ridgelet transform domain can ensure the fidelity of the covered image. Attack experiment results show that compared to the wavelet transform domain watermarking algorithm, the proposed algorithm can not only resist the common attacks such as compression, noise, filtering and cropping, but also resist the geometric rotation attack.