一种用于篡改定位及近似恢复的水印算法

Watermarking Method with Tamper Localization and Approximate Recovery Capability

针对遥感影像易被非法篡改的问题,提出了一种用于篡改定位及近似恢复的遥感影像半脆弱水印方案。首先将二值伪随机矩阵作为认证水印,通过量化Contourlet变换各方向子带中绝对值最大的系数将其嵌入,然后采用影像4×4分块后的灰度平均值作为恢复水印,将其嵌入到最低有效位。水印检测不需要原始遥感影像参与,实现了盲提取。实验结果表明,算法对于噪声、JPEG压缩等合理性失真具有一定的鲁棒性,而对于恶意篡改等攻击具有脆弱性,能够实现篡改区域的精确定位和近似恢复,使恢复后的遥感影像满足基本的视觉要求,可有效保护遥感影像的安全。

A semi-fragile watermarking method with tamper localization and recovery capability was proposed in order to settle the problem that remote sensing images were easily to be tampered maliciously. Authentication watermarking, composed of a binary random matrix, was embedded through the way of quantizing the coefficients with maximum absolute value in directional subbands of Contourlet transform. Recovery watermarking, which was formed by the average gray value of every non-overlapping 4＆4 block, was embedded in the least significant hit. The watermarking detection was a blind extraction procedure, which could be implemented directly without using the host images. Experimental results demonstrated that the method could resist adding noise, JPEG compression and other rational distortions to a certain extent. Meanwhile, it was fragile to malicious manipulation, and couhl lo- cate the tampered regions accurately and recover them approximately, and the recovered images could meet the basic visual need. So the proposed semi-fragile watermarking method was able to ensure images security effectively.