基于Hough变换旋转校正的QR码全息水印算法

目的为了解决QR码图像不易隐藏水印的难题,提出一种针对QR码载体图像的强鲁棒性频率域数字水印算法。方法利用经双随机相位调制的傅里叶全息技术对水印进行加密,同时对QR码载体图像进行包含高斯模糊以及去除视觉敏感区域的预处理,将全息加密水印的奇异值矩阵加权嵌入到经过二级小波分解的QR码的中频系数后进行还原。Hough变换通过坐标系转换检测QR码图像同方向边缘的直线,后进行旋转校正,以提取水印。结果在合适的嵌入强度下,水印保持了良好的不可见性(PSNR=32.3)。经过二值化攻击、信号处理(滤波、噪声、压缩)、几何变换(平移、旋转、缩放)以及打印扫描处理等一系列攻击,提取出的水印依旧清晰可见。结论提出的算法针对各类攻击有较强的鲁棒性,可抵御任意角度旋转攻击,同时兼顾水印的不可见性,针对QR码的防伪技术具有一定的应用价值。

Algorithm of QR Code Holographic Watermark Based on Skew Correction by Hough Transform

The work aims to put forward a frequency domain digital watermarking algorithm for QR code carrier image which has strong robustness, in order to overcome the problem that QR code image is hard to hide watermarks. Fourier holography technology subject to double random phase modulation was used to encrypt watermark and the pretreatment was made including Gaussian blur and the removal of visual sensitive area for QR code carrier image at the same time. Then the singular value matrix of holography encrypted watermark was weighted and embedded into the middle frequency coefficient of QR code after secondary wavelet decomposition and then was restored. The straight line at the edge of QR code image in the same direction was detected by Hough transform through coordinate system transformation and then skew correction was done to extract the watermark. Watermark kept good invisibility (PSNR=32.3) under the suitable embedding strength. After a series of attacks, including binarization attack, signal processing (filtering, noise and compression), geometric transformation (translation, rotation and scaling) and print-scan attack, the extracted watermark was still clearly visible. The proposed algorithm has strong robustness for various types of attacks and it can resist rotation attack at any angle, and it also meets the requirement of invisibleness. The algorithm has certain application value in QR code anti-counterfeit technology.