一种自适应空间域图像数字水印算法

Adaptive Spatial Domain Digital Watermarking Algorithm for Image

为解决空间域水印算法鲁棒性较差,水印提取复杂等问题,提出了一种自适应空域数字水印算法。建立了BP神经网络模型,通过学习、训练记忆每个图像块中像素之间的关系,可以实现嵌入强度的自适应调整以及水印盲提取。详细论述了水印嵌入和提取算法,对水印图像进行伪随机置乱,可以减少对宿主图像像素点的破坏,提高了提取水印与原始水印的相关度;利用对比度函数确定水印嵌入位置,并得到密钥信息用于水印提取。实验验证,峰值信噪比(PSNR)可以达到57;在JPEG压缩、噪声、剪切和旋转等常见的图像攻击下,归一化相似度(NC)均大于0.9。表明了所述算法的不可感知性和抗攻击性能。

In order to solve the problems of spatial domain watermarking algorithm such as poor robustness,watermarking extraction complex and so on, an adaptive spatial domain watermarking algorithm is proposed. The BP neural network model is established. The adaptive adjustment of watermarking embedding intensity and its blind extraction can be realized by the relationship between the pixels in each image block. The watermarking embedding and extraction algorithm are described in detail.The damage to the host image pixels can be reduced as well as the relevance of the extracted and original watermarking is also improved by watermark image pseudo-random permutation. Contrast function is used to determine the watermark embedding positions,and the key information is obtained for watermark extraction. The experimental results show that peak signal noise ratio( PSNR) can achieve to 57. The normalized similarities( NC) are greater than 0. 9 under the common image attacks such as JPEG compression,noise, cutting and rotating. The algorithm described has imperceptibility and anti-attacking.