Research On Color Image Watermarking Algorithm With Quaternion Visual Features

With the advancement of 5G technology and the rapid popularization of personal mobile terminals,digital multimedia products such as images and videos are facilitated the dissemination,and at the same time the risk of tampering and forgery of original data has also increased.The issue of protecting data content security and copyright authentication is becoming more serious.The emergence of digital watermarking technology provides a solution to this information security problem.In recent years,digital watermarking technology has developed rapidly,but there is still a contradiction between invisibility and robustness in watermarking algorithm.Based on the study of human visual characteristics,researchers have put forward Just Noticeable Distortion(JND)model to guide watermarking algorithm,which provides a way of balancing the contradiction between invisibility and robustness.Fusion of visual models,the watermark embedding is more consistent with the human visual characteristics,so that the watermarked image maintains a high image quality under the premise of improving the robust performance of the algorithm.For the current watermarking algorithm proposed for color images,the JND model is often only applied to the brightness channel after the color model conversion,or the traditional monochrome model is used to process the three color channels of RGB separately,which ignores the connection of the three color channels of color images.Therefore,this article applies the quaternion algebra theory to describe the color image as a whole,expands the study of color images to the quaternion space,studies the visual characteristics of the image in the quaternion domain,and designs the corresponding visual perception model,combined with Spread Transform Dither Modulation(STDM)embedding scheme,a robust quantized watermarking algorithm is proposed.The main contents are as follows:1.In order to solve the problem that the color image watermarking scheme ignores the connection and integrity of three channels,the quaternion theory is applied to represent the color image as a four-dimensional spatial matrix,and the Quaternion Discrete Cosine Transform(QDCT)is used to process the color image.The influence of the transformation parameter ? in the QDCT transformation process is studied,and the perceptual unit pure quaternion is proposed to improve the performance of the watermarking algorithm.2.A quaternion domain JND model for spatial masking,contrast masking and brightness adaptive features of color images is established based on human visual perception characteristics.At the same time,the QDCT channel coefficient components are used to extract color information effectively,and the chromaticity perception masking factor is obtained.Combined with the above visual perception features,a robust visual perception JND model in quaternion domain is established.3.Based on the quaternion domain visual perception JND model,combined with the traditional STDM watermarking algorithm,a color image watermarking algorithm with visual features is designed.Experimental results show that,compared with the existing watermarking schemes,the proposed algorithm has stronger robustness under the premise of uniform image perception quality.