Research On Reversible Watermarking Based On Human Visual System

The rapid spread use of Internet and the wide application of intelligent terminal bringpeople great convenience on life, work, and study, but it also creates challenges. Mediainformation can easily be illegally copied and manipulated in an open networkenvironment. Therefore, digital copyright protection attracts wide attention around theworld. Digital watermarking technology provides effective solutions for such problems,and it will become perfect gradually. Traditional digital watermark techniques bring impacton quality of the images after watermarking embedding and extracting. But some specialapplications require high quality of the image, such as military and medical domain. Thesekinds of images can not tolerate any distortions. So reversible watermarking is proposed tosolve this problem, which not only extracts the watermark, but also can restore the originalimage without distortion.This paper first introduces the background of watermarking technology, applicationdomain, classification and characteristics. Then the reversible watermarking is researchedand discussed, especially the reversible watermarking algorithm about characteristics ofhuman eyes. The main research contents of this paper are as follows:(1) A novel reversible grayscale watermarking algorithm based on human visualsystem, integer wavelet transforms and multi-objective genetic algorithm is proposed.Firstly, integer wavelet transform is performed on the image, and then the coefficients arepartitioned into blocks. In each local block two appropriate pixels are found through pixelgray value. According to modulo between the gray value difference of the two pixels andsix, the third appropriate pixel is found. Different with other reversible algorithm,watermark is embedded into image by modifying the three gray values of the originalimage. Human visual system is used to improve invisibility. The algorithm calculates watermarking embedding strength dynamically by normalizing entropy of each block.Finally, genetic algorithm is used to optimize results, which meets the quality andinvisibility demand after the watermark embedding.(2) In order to ensure invisibility and improve robustness of the watermarked image, arobust reversible watermarking algorithm is proposed. Statistic is used to increase the robustof the reversible watermarking. Firstly, the image is cut into several blocks and appliedinteger wavelet transforms to them. The middle point of each block is chosen as dividingpoint, and divides the block into two sections. Then the mean values of each part arecalculated. Watermarking is embedded into the host image by modifying the gray values ofthe dividing points. According to the two obtained mean values of each block and thewatermark, the middle point’s gray value is set up in different scopes. At the same time,histogram modification is performed on the original image to prevent overflow andunderflow and the characteristics of an image should be taken into consideration inselecting blocks to be embedded. The sum of visual edge and feature entropy of each blockis calculated, and then the values are sorted by magnitude. The blocks with low values areselected as the candidates for embedding, which improves the imperceptibility androbustness of watermarking.