基于随机分布圆形模板的抗屏摄盲水印算法

针对现有的抗屏摄水印算法不能兼顾鲁棒性、不可见性和水印容量等问题,提出一种基于随机分布圆形模板的抗屏摄鲁棒盲水印算法。首先将十六进制水印序列转换为固定的角度值,再根据随机线条分布算法生成一系列带有角度信息的圆形模板,将原宿主图像蓝色分量进行整数小波变换并取出低频子带（LL）,对LL分块并用密钥选出若干水印嵌入块,最后将圆形模板嵌入被选出的块中。另外,为了提高水印容量,选择十六进制水印密码以代替二维码或条形码;为了提高水印不可见性,将水印模板嵌入整数小波变换的低频子带;为了平衡水印透明性和鲁棒性,利用粒子群寻优算法（PSO）自适应地得出最佳嵌入强度;为了解决拍摄后造成的图像畸变,在水印提取前对拍摄的图像进行几何校正。在数据集USC-SIPI上选取了测试图,进行了常规攻击与屏摄攻击实验,并与该领域主流的六种方法进行了比较。实验结果表明,在保证不可见性的前提下,该算法能抵抗各种常规类型攻击。在一定范围的拍摄距离和拍摄角度内,算法能有效提取出水印密码,并提高了水印容量,在抗屏摄方面具有良好的鲁棒性。

Anti-screening blind watermarking algorithm based on randomly distributed circular template

Aiming at solving the problems that the existing anti-screening watermarking algorithms cannot take into account the robustness, invisibility and watermark capacity, this paper propose a robust blind watermarking algorithm based on randomly distributed circular templates. Firstly, it converted the hexadecimal watermark sequence into a fixed angle value, and then generated a series of circular templates with angle information according to the random line distribution algorithm. Secondly, it separated the original host image according to the three primary color channels and take out the blue component. Thirdly, it performed integer wavelet transform on blue component and took out low frequency sub-band(LL). Fourthly, it used divided the LL into blocks and select several watermark embedding blocks with a selected key. Finally, it embeded the circular templates in the selected blocks. In addition, in order to increase the watermark capacity, it choosed the hexadecimal watermark password to replace the QR code or barcode. In order to improve the invisibility of the watermark, it embeded the watermark templates in the low frequency sub-band of the integer wavelet transform. In order to balance the transparency and robustness of the watermark, it used the particle swarm optimization algorithm(PSO) to obtain the optimal embedding strength adaptively. In order to solve the image distortion caused by shooting, it performed geometric correction on the captured image before watermark extraction. This paper selected images on the data set USC-SIPI to implement experiments related to common attacks and screen capture attacks, and compared the proposed algorithm with other six algorithms with this image data set. The experimental results show that the algorithm can not only ensure the invisibility of watermark but also resist various types of common attacks. And within a certain range of shooting distance and shooting angle, the algorithm can effectively extract the watermark password, it also improves the watermark capacity and has good robustness against screen shots.