基于元胞自动机的自嵌入图像认证水印算法

Cellular automata-based self-embedding watermark scheme for image authentication

提出了一种用于图像认证的分块自嵌入水印算法,图像分块后,根据基密钥产生的基函数,对子块进行两层元胞自动机变换,由高层低频系数产生初始构型,再由整数混沌序列计算对应子块的局部规则,由该局部规则的元胞自动机从初始构型演化生成基于图像内容的水印,再用量化系数技术将水印嵌入低层低频系数,利用元胞自动机对初始构型和局部规则的多变性,能有效检测并定位对图像内容的恶意篡改,避免矢量量化等攻击,同时也增强了算法的安全性,实验结果表明算法对JPEG压缩有较好的鲁棒性,检测时不需要原始图像。

A self-embedding scheme for image authentication is presented. The scheme splits the original image into blocks and performs two layer cellular automata transform on blocks with a transform base function produced by a base-key. An initial configuration is produced using high-layer low frequency coefficients, and a local rule of a block is calculated from a chaotic sequence. The watermark is generated from the initial configuration through an elementary cellular automaton of the local rule and embedded into the low-layer low frequency coefficients with quantization technique. The proposed scheme is exploiting the flexibility on initial configuration and local rule can localize the tampers on the watermarked image, and resist the counterfeiting attacks on block-wise independent watermarking schemes, moreover increase the watermark security. Experiment results show the proposed scheme maintains certain robustness for JPEG compression and no host image is needed in watermark detection.