Research On Zero Watermarking Of Color Image Based On Tensor

Driven by powerful science and technology,today's society can be said to be changing rapidly,and all kinds of information carriers have modern forms of expression and processing.As a relatively more visual and intuitive data carrier,image mainly exists in digital form at present.However,in this case,criminals can use the convenient network and some software to download,copy,and tamper with digital images,which seriously threatens the legitimate rights and interests of copyright owners.Digital zero watermark technology can effectively protect the copyright of digital image,mainly by extracting the feature data of the original carrier image to construct a unique zero watermark.But the existing algorithms are mainly applied to grayscale images.Considering the above factors,a new digital zero watermarking algorithm based on color image is proposed in this paper.Since the image used is a three-dimensional color image,tensor and tensor expansion are introduced into the algorithm,which is almost the first attempt of tensor concept and related technologies in the field of zero watermark.During the operation,first take the data of each dimension from the original carrier image,construct different three-dimensional tensors,and then perform tensor expansion on the obtained tensors according to a specific direction.These several steps have achieved three functions at the same time,including the full use of all aspects of the color image data,lossless dimensional compression and data scrambling.In addition,in the subsequent process,singular value decomposition(SVD)and discrete cosine transformation(DCT)are also used to process different data after the tensor expansion.The above two operations can reduce the amount of zero watermark data and promote the improvement of the robustness.According to the experimental results of a large number of simulation experiments and comparison experiments,it can be determined that the algorithm proposed in this paper has both good uniqueness and strong robustness.Moreover,when the algorithm is applied to different types of original carrier images,the performance of the algorithm is stillmaintained at a good level.At the end of the paper,based on some shortcomings of the algorithm in this paper,the further research directions are roughly explained.