基于纹理复杂度的图像篡改定位和恢复水印算法

A tamper location and restoration approach for watermarking using image texture complexity

针对现有方法对图像纹理相关性利用不充分导致图像块特征值选取不灵活和隐蔽性不足的问题,提出了一种新的基于纹理复杂度和Laplacian算子的篡改定位和恢复水印算法。算法首先将图像分成3×3的分块,分别计算各块的纹理复杂度,利用纹理复杂度对同质块进行合并,降低水印嵌入对图像质量的影响;然后,计算母块和单独块的拉普拉斯卷积值作为其特征值,并将母块的位置信息作为子块的特征值,新的特征值增强了对篡改的敏感性,并减少了需嵌入的特征水印数量;最后将子块的特征值水印嵌入到母块像素最低有效位中,母块和单独块则嵌入到其对应块的像素最低有效位中。在水印图像没有受到篡改时,水印图像平均PSNR较现有方法提升7.83 dB左右;当受到篡改时,恢复图像质量提升1.01 dB左右。实验结果表明,本文算法能够增强图像水印隐蔽性,提高篡改图像恢复质量,适用于版权保护和医学图像的篡改检测。

Aiming at solving the existing problem that the insufficient utilization of image texture correlation causes the eigenvalues of image blocks to be selected inflexibly and the concealment is insufficient,a new tamper localization and restoration watermarking algorithm based on texture complexity and Laplacian operator is proposed.The algorithm first divides the image into 3×3 blocks,and the texture complexity of each block is calculated respectively,according to texture complexity to merge the homogenous blocks to reduce the influence of watermark embedding on the image quality.Then the Laplacian convolution of the mother block and the individual blocks is calculated as their eigenvalues,and the position information of the mother block is taken as the eigenvalue of the sub-blocks.The new eigenvalue enhances the susceptibility to tampering and reduces the number of feature watermarks to be embedded.Finally,the eigenvalue watermarks of sub-blocks are embedded in the least significant bit of the mother block′s pixel,and the mother block and the individual block are embedded in the least significant bit of the pixel in its corresponding block.When the watermark image is not tampered,the average PSNR of the watermark image is improved by about 7.83 dB compared with the existing method.When it is tampered,the restored image quality is improved by about 1.01 dB.The experiments illustrate that the proposed algorithm can enhance the image watermark concealment and improve the quality of tampered image recovery.It is suitable for copyright protection and medical image tamper detection.