为了弥补空间域水印算法在提取水印时需要原宿主图像、鲁棒性较差等缺点, 提出一种基于BP神经网络和人眼视觉特性(HVS)的空域数字水印算法。先选出原宿主图像对比度函数值最大的前块用于二值水印图像的嵌入, 利用神经网络能够逼近任意非线性关系的特点和其自适应性, 构建水印信号嵌入前后图像块像素值间的映射关系, 实现了水印的盲提取, 并将该算法与现有算法进行了对比。实验结果表明, 对于JPEG压缩、加噪、剪切和旋转等常见的图像处理攻击, 算法具有良好的鲁棒性和不可见性。

To counter the shortcomings of spatial domain watermarking algorithms such as poor robustness, needing the original host image when extracting the watermark image, a spatial domain image watermarking algorithm based on BP neural network and human visual system (HVS) is proposed. The first pieces of the original host image with the largest value of contrast function are selected to perform embedding of binary watermarking image. The relationship between pixels value before and after the watermark signal embedded is established, taking advantage of the attributes of the neural network such as nonlinear approximation and strong learning ability, thus realizing the blind watermark extraction, which is compared with other watermarking algorithms available. The experimental results show that the algorithm put forward here has good robustness and imperceptibility against common image processing attacks such as JPEG compression, noise, cropping, rotation etc.