Grayscale Watermarking Algorithm Combining Texture Complexity And BEM

With the development of information media technology,the transmission of multimedia information such as digital images and videos has become very convenient.Due to the lack of people’s awareness of network copyright,too convenient transmission has caused to many multi-medias being copied and modified without authorization,which greatly affects the development of digital network information.To solve these problems,the concept of digital watermarking was proposed.In recent years,digital watermarking has developed rapidly,but there are still many problems.Most watermarking algorithms have the limitation of resisting a specific or a special type of attack,so they have low robustness under most attacks,even the watermark information with representative identity can’t be extracted.Gray image watermarking has become a research hotspot of digital watermark because of its large amount of stored information and more vivid expression of image information.However,the more complex pixel gray value also leads to the reduction of invisibility and robustness of watermark.How to balance the invisibility and robustness of watermarking algorithm has become a research hotspot of digital watermarking at home and abroad.In order to improve the robustness of gray image watermarking algorithm against rotation,large-scale clipping,compression and noise,and reduce the invisibility of embedded watermark,a combination of texture complexity and two-dimensional empirical mode decomposition(BEMD)is proposed in this paper.The specific work is as follows:(1)In the encryption process of the watermark image,Arnold transform,which is easy to operate and easy to recover,is used to scramble the watermark image to improve the security of it.Then the watermark image is decomposed by BEMD.Through the decomposition,it is found that the loss of watermark information at different levels has different effects on the watermark image,so different embedding parameters are set.For the less important layers of the watermark,the embedding coefficient is reduced.Thus,the invisibility of some watermark information is reduced.(2)By analyzing the texture complexity of the host image,an appropriate amount of regions with rich texture are selected,and the watermark information of different scales after BEMD decomposition is embedded into the same scale of the texture rich region of the corresponding scale.According to the masking of human vision,the region is not easy to be recognized by human eyes,so as to improve the overall invisibility of the watermark.(3)After the BEMD decomposition of the host image under the same conditions,multi-layer IMFs with different scales are obtained,which have sufficient embedding capacity for the watermark information and sufficient embedding space for the gray watermark image with rich embedded information.The watermark information is repeatedly embedded,screened and voted when extracting the watermark,which improves the resistance to large-scale shear attack,rotation attack JPEG compression attacks and high-intensity salt and pepper attacks.(4)For the block embedded watermarking algorithms,there will be blocking effect and edge loss during frequency domain conversion.In this paper,the host image is globe decomposed by BEMD first,and then the decomposed IMF and residual are divided into blocks,so as to effectively avoid the edge effect caused by the decomposition of the host image after blocking.By embedding different watermark images into different host images,it is obtained that the peak signal-to-noise ratio of the embedded watermark image is more than 40 d B,and the mean NC values under eight common attacks is more than 0.95.Compared with the existing algorithms,the proposed algorithm performs well in large-scale clipping,noise attack and filtering attack.The experimental results are better than the comparison algorithm in other attacks.The proposed algorithm has good invisibility and robustness...