CAJ | 학술논문

为了解决现有数字水印中鲁棒性和不可感知性之间的矛盾，设计一种组合非负矩阵分解和奇异值分解的鲁棒图像零水印算法。首先对原始图像进行不重叠分块，分别对每子块图像进行秩为r的非负矩阵分解；然后对非负矩阵得到的系数矩阵作奇异值分解得到特征向量；最后将特征向量和水印的运算得到代表原始图像版权信息的零水印序列。实验结果表明该方案在有效降低算法复杂度的同时对常见信号处理具有很强的鲁棒性。
위료해결현유수자수인중로봉성화불가감지성지간적모순，설계일충조합비부구진분해화기이치분해적로봉도상령수인산법。수선대원시도상진행불중첩분괴，분별대매자괴도상진행질위r적비부구진분해；연후대비부구진득도적계수구진작기이치분해득도특정향량；최후장특정향량화수인적운산득도대표원시도상판권신식적령수인서렬。실험결과표명해방안재유효강저산법복잡도적동시대상견신호처리구유흔강적로봉성。
In order solve the contradiction between the robustness and imperceptibility in current watermarking , we design a robust image zero-watermarking algorithm in this paper which combines the non-negative matrix factorisation ( NMF) and the singular value decomposition (SVD).First the original image is divided to non-overlapping image blocks, and the NMF of rank r is applied to each sub-block image;Then the SVD is applied to the factorised coefficients matrix to obtain the eigenvector; Finally the zero-watermarking sequence representing the copyright information of original image is derived through the calculation of the eigenvector and watermark .Experimental results show that the scheme has strong robustness on common signal processing while effectively reduces the computation complexity .