提出一种基于重叠正交变换的奇异值分解图像水印方案,应用量化法嵌入信息,在提取中无需原始图像和水印信息。由于重叠正交变换能有效去除基于块数据变换所产生的块效应,而图像的奇异值具有稳定性,它体现图像的内蕴特性而非直接的视觉特性,因而本文方法具有良好的鲁棒性和不可见性。
A digital watermarking method based on lapped orthogonal transform (LOT) and singular value decomposition (SVD) is proposed. LOT can reduce blocking effects caused by other block-based transforms such as DCT. Sigular values represent inherent characteristics of the image and are therefore quite stable. By using a quantization scheme, the embedded watermark can be detected without the cover image and the watermark. Experiments show that the embedded watermark is transparent and robust agaist a number of attacks.