一种基于视觉模型的小波域图像盲水印算法

A Blind Watermarking Algorithm Based on DWT and Visual Models

本文提出了一种基于混沌置乱和JND阈值的小波域图像盲水印算法。依据人眼的视觉特性,对宿主图像进行离散小波变换,在小波域子带LH2、HL2幅值较大的系数以及LL2幅值较小的系数中嵌入水印,水印的嵌入采用量化调制的方式,并取JND阈值作为量化间隔,使水印实现盲提取。实验结果表明,在一定不可见性的情况下,此算法对一些常见的水印攻击有较好的鲁棒性。

The work concerns with a new blind watermarking algorithm based on JND and chaotic scrambling on discrete wavelet transform （DWT）. Based on the visual masking effect of the human visual system, the proposed algorithm embeds watermarks into some big coefficients of sub-bands LH2, HL2 and some small coefficients of sub-band LL2 in the discrete wavelet domain of an image. In order to extract the watermarks without using the original image, we insert the watermarks by using quantized modulation and take JND as the quantized interval. The experimental results demonstrate that the proposed algorithm is invisible and robust to certain common distortions.