Research On Digital Watermarking Algorithm Combining Global And Local Texture Features

With the advent of 5G and the rise of various image sharing platforms and APPs,the spread of digital images is faster and more convenient,and at the same time the cost of piracy is also reduced.As an effect solution for prevent piracy,digital watermarking has received increasing attention from researchers.Digital watermarking technology refers to embedding watermark information such as copyright into digital image and other carriers.After the digital watermark is embedded,the visual quality of the image will not change.In the event of infringement,copyright protection can be achieved by detecting watermark information.In order to improve the fidelity and robustness of watermarking algorithms,researchers have introduced Just Noticeable Difference(JND)models to digital watermarking algorithms.However,the traditional JND model does not consider the masking effect of texture regular features on the human visual system when simulating human perception characteristics.Therefore,it is impossible to estimate the true visual redundancy well.In this paper,the human visual system's perceptual characteristics of image texture features are fully considered.The visual saliency map of the image is used to calculate the texture regularity characteristics of the image block,and the texture regularity factor is used as the modulation factor of the JND model.This paper applies the proposed visual JND model to a digital watermarking algorithm,which enables the watermarking algorithm to calculate the visual redundancy more accurately,and improves the robustness of the watermarking algorithm while maintaining the image quality.The texture feature of an image refers to the repetition and arrangement order of the elements contained in the image,which can reflect the surface features of the image or things in the image.The texture feature of an image is not a feature of a certain pixel,but a feature obtained by all pixels in a certain area,which has strong anti-interference ability and rotation invariance.Most of the traditional texture regularity measurement methods are directly calculated in the pixel domain and the spectral domain,without considering human perception characteristics.This paper proposes a method for measuring the texture regularity based on human visual characteristics,and solves the problem of measuring the texture regularity when the saliency map does not match the image size.The main research results of the paper are:1.In order to solve the problem of measuring the texture regularity of the image when the size of the visual saliency map is different from the original image,a method of measuring the texture regularity of the image based on the saliency map of any size is proposed.By analyzing the texture regularity in each area of the visual saliency map,the regional regularity score is obtained,and then the texture regularity measure of the entire image is obtained.The experiment proves that the proposed method can reflect the texture regularity of the whole image well.2.Aiming at the sensitivity of the human visual system to the texture regularity of images in watermark processing,this paper proposes a block-based calculation method of texture regularity and uses it as a factor to form a JND model.As an important element of the JND model,this factor combines the texture regularity factors within and between image blocks,which can more effectively describe the visual characteristics of the human visual system,making the proposed JND model more robust.3.In order to improve the robustness and invisibility of the digital watermarking algorithm,the proposed visual JND model is applied to the image digital watermarking algorithm to obtain a digital watermarking algorithm combining global and local texture features.Experiments show that this method can more accurately analyze the visual redundancy of the human eye and improve the robustness of the watermark while maintaining the fidelity of the image.