Research On Digital Watermarking Algorithm Based On Quantization And Compressive Sensing

With the rapid development of Internet technology,multimedia technology is rising rapidly.The production,storage,and dissemination of multimedia data,including text,audio,image,and video,become more and more convenient and fast without leading to the decline of content quality.However,the multimedia data transmitted through the network are easy to be illegally copied,modified,and disseminated by pirates,which makes the rights and interests of copyright owners are not guaranteed.Digital watermarking technology can protect copyright and prove its reliability by adding some information into multimedia data.To better realize the copyright protection and content authentication of digital image,this paper studies robust image watermarking algorithm,semi-fragile image zero-watermarking algorithm and image dual watermarking algorithm upon the foundation of the quantization watermarking algorithm and the compressive sensing theory.The main works are as follows:1.In order to achieve a better trade-off between the imperceptibility and the robustness of watermark,this paper studies a modified logarithmic quantization index modulation watermarking method based on visual saliency and contourlet transform.In this algorithm,salient object detection and contourlet transform are operated separately on the host image.After dividing the transformed low-pass sub-band into nonoverlapping blocks,the quantization stepsize of each block is determined by the saliency values and the energy distribution of the block.The logarithmic quantization index modulation is modified in order to achieve the robustness against the amplitude scaling attack.Then,the watermark bits are embedded into the maximum singular value of the low-pass sub-band block using the modified method.Experiment results demonstrate that the proposed method has satisfactory imperceptibility and is more robust to different types of attacks compared with LQIM and other algorithms.2.In order to further improve the robustness and the security of watermark,a novel angle quantization index modulation watermarking scheme based on block compressive sensing is proposed.The sparse random matrices constructed by chaotic sequence are chosen as the measurement matrix and the block compressive sensing is carried out on the host image.In each block,two optimal measurements are chosen to embed the watermark using AQIM.Then,the watermarked image is recovered through Min-TV reconstruction algorithm.Experiment results demonstrate the security of the algorithm and show that the proposed method outperforms common existing methods in terms of the robustness.3.Aiming at improving the accuracy of image tampering detection,this paper proposes a semifragile zero-watermarking algorithm in DWT domain based on block compressed sensing.DWT is applied to the test image and then four sub-bands are carried out block compressive sensing,seperately.The measurements of four sub-bands corresponding to the same image block are combined as the zero-watermarking information of the block.In step of detection,only the LL sub-band measurements in zero-watermarking is used to detect and locate tampering.If the tampered image block is detected,the whole zero-watermarking of the block is used to restore the original image block.The experimental results show that the algorithm can detect and recover tampered images effectively,and its performance is improved compared with the BCS-SFZ algorithm.4.To aim at solving the problem of single function of watermarking algorithm,this paper proposes a dual watermarking algorithm upon the foundation of angle quantization index modulation watermarking algorithm and semi-fragile zero-watermarking algorithm in DWT domain based on block compressed sensing.In the algorithm,robust watermarking is embedded firstly.Then,DWT is applied to the watermarked image and the semi-fragile zero-watermarking is constructed.When detecting and extracting watermark,the semi-fragile zero-watermarking is used to judge whether the image has been tampered.If it has been tampered,the untampered image is restored by zero-watermarking,and then the robust watermarking is extracted;Otherwise,robust watermarking can be extracted directly.The experimental results show that the algorithm not only maintains the good anti-attack ability of robust watermarking,but also detects malicious tampering and recovers tampered areas effectively.