数字图像的自适应公开水印技术

该文提出一种以灰度层图案作水印的自适应公开水印（public watermarking)技术,这种技术在不可感知性和稳健性等方面面临着更大的挑战,而探索图像新的稳定性特征并结合人类视觉系统（HVS）的性质来设计水印算法正是克服这一难题的有效途径。为此,该文分析了一种新的变换域－多小波（multiwavelet)域系数的特征,并将基于单小波域的人眼噪声敏感性模型推广到多小波域,在此基础上,先将灰度层图案水印分解为二值位图平面,然后自适应地将各个位图平嵌入多小波变换域中,实验结果表明,该文方法对多种图像处理具有很好的稳健性。其性能优于文献上的相近方法。

Adaptive Public Watermarking for Digital Images with a Gray-Level Watermark

For improving robustness, exploring stable characters of an image and taking into account characters of the HVS are very effective strategies. Based on these ideas, this paper proposes a novel public watermarking method in which a gray-level logo watermark is embedded into multiwavelet domain. Like the scalar wavelet case, the multiwavelet decomposition provides us a multiresolution representation of an image, however there are four different "detail" coefficient subblocks in the same direction and in the same resolution. It is found that there are two characters in the multiwavelet domain. One is that values of corresponding pixels in two of subblocks in the same direction and resolution are almost equal, and another is that the polarity between them is very stable under common image processing operators. According to these unique characters, the gray-level watermark is firstly decomposed into bitplanes, and then each bitplane is embedded into multiwavelet domain by directly modifying the polarity between them. Moreover, in order to embed the watermark into an image the maximum, are exploited the characters of the HVS. Here, the noise sensitivity model of the eye is extended based on wavelet domain proposed by Lewis and Knowles so that it can be suitable for the multiwavelet domain case. Experimental results show that the proposed technique has good robustness to common image processing operators, and also is superior to the similar method in the literature.