一种基于压缩感知的无损图像认证算法

Lossless image authentication algorithm based on compressive sensing

针对图像无损认证和篡改检测问题,提出了一种基于压缩感知的自嵌入水印算法.首先,将图像划分为若干分块;然后,利用压缩感知理论对各图像块进行观测,将得到的观测值作为图像内容特征,通过基于哈希的消息认证码算法生成图像摘要,并将摘要信息作为水印嵌入到图像小波域中,嵌入方式采用量化索引调制算法.在认证端提取水印并恢复图像,然后对图像进行压缩感知随机投影,将得到的摘要信息与提取的水印进行对比,实现图像认证和篡改检测.仿真实验结果表明,水印嵌入后图像的峰值信噪比可达到39 dB以上,说明该算法具有较好的不可感知性和隐蔽性.此外,该算法还具有较强的局部篡改定位性能,图像认证块大小可根据图像进行相应调整.若图像通过认证,则可以无损地恢复原始图像.

A self-embedding watermarking algorithm based on compressive sensing is proposed for lossless image authentication and tamper detection.First,the image is divided into several sub-blocks.Secondly,each image block is observed by using the compressive sensing theory.Then,the compressive measurements,which are taken as the content features of the image,are used to generate the image digest by the hash-based message authentication code algorithm.Finally,the image digest is embedded into the wavelet coefficients as the watermark by the quantisation index modulation algorithm.During authentication,the watermark is extracted and the image is recovered.Then,random projection is conducted on the recovered image to generate the digest.By comparing the digest with the extracted watermark,image authentication and tamper detection are implemented.The simulation results show that after embedding watermark the peak signal-to-noise ratio is above 39dB,proving the imperceptibility and elusiveness of the proposed algorithm.Besides,the algorithm can accurately localize the tamper.The size of the authentication block can be adjusted according to the image.Once the image passes the authentication,the original image can be recovered without any loss.