一种准确定位篡改的遥感图像盲脆弱水印算法

A high-accuracy tampering localization algorithm for remote sensing images using fragile watermarking

针对现有块级脆弱水印方案的虚警问题,该文提出了一种准确定位篡改的遥感图像盲脆弱水印算法。首先将图像分解为不重叠的2×2块序列,接着利用逻辑映射和块自身信息进行多重异或操作,生成8位脆弱水印,最后在每个块内进行2位LSB替换,实现脆弱水印的嵌入。另外,还提出了一种虚警优化策略,可以有效减少虚警,提升定位准确度。实验结果表明,所提算法具有较好的图像视觉质量,可同时适用于8位和16位的单波段和多波段遥感图像。当面对复制粘贴,内容消除,添加虚假版权以及剪切攻击时,可实现较好的定位效果,尤其在应用于16位遥感图像上时,定位准确度可达到99%以上,优于其他算法。该文所提算法可为互联网云环境下遥感图像的安全保护提供技术保障。

Given the false alarm problem in existing block-based fragile watermarking schemes,a blind fragile watermarking algorithm for accurately locating tampering in remote sensing images is proposed.Firstly,the cover image was decomposed into a non-overlapping sequence of 2×2blocks,and then an 8-bit fragile watermark was generated by performing multiple XOR operations using logical mapping and block self-information.Finally,the fragile watermark was embedded by replacing the 2least significant bits(LSB)in each block.In addition,a false alarm optimization strategy was proposed to effectively reduce false alarms and improve localization accuracy.Experimental results showed that the proposed algorithm had good visual quality and could be applied to both 8-bit and 16-bit single-band and multi-band remote sensing images.When facing copy-paste,content removal,false copyright addition,and cropping attacks,the proposed algorithm achieved good localization performance,especially for 16-bit remote sensing images,with a localization accuracy of over 99%,which was better than other algorithms.This algorithm could provide technical support for the security protection of remote sensing images in the Internet cloud environment.