Study On Reversible Watermarking Based On Pixel Value Ordering

With the advances in generation,storage,and communication technology of digital content,the security of digital information has received more and more attention in the past two decades.Digital watermarking is one of the effective solutions for tamper detection and digital content protection.Due to the extra ability to recover the host media exactly as well as extracting the embedded data,reversible data hiding(RDH)has become a most important branch of data hiding and has been applied to many quality sensitive fields such as medical imaging,remote sensing and military.Among the existing reversible watermarking schemes,the pixel value ordering(PVO)based schemes have achieved excellent performance in small embedded capacity.In this method,only the pixels with the largest and smallest value in the block are predicted and modified to embed the data.The preservation of the sequence of pixels guarantees the reversibility.In addition,the complexity of pixel blocks is measured to achieve a higher performance.By researching the existing RDH schemes,this thesis proposes a gradient-complexity-based pixel value ordering(PVO)scheme.In this scheme,the image has been divided into blocks with two different sizes.Then calculate the gradient complexity of the two kind of blocks separately.Using larger block's complexity data to help sorting the small one to optimize the embedding order.Then,by calculating the gradient complexity of each block,different embedding block sizes are dynamically selected.The lower the gradient complexity has,the smoother the area where the block is located,so use smaller blocks to increase the embedded capacity.Blocks located in rougher areas use larger blocks to ensure the prediction accuracy.Finally,by constructing the PAM(Prediction Accuracy Matrix)to iterate the optimal combination of the embedding thresholds under different embedding capacity,the complexity of the algorithm is greatly reduced.The proposed method in this thesis is a reversible data hiding scheme for grayscale images.Through experimental simulation,it is verified that the algorithm has significant performance improvement compared to the existing algorithms under different embedding capacity.Through the statistics of the running time between the proposed method with the traditional schemes,the optimization of the proposed algorithm on the time complexity is demonstrated.The time consuming of threshold iteration in dynamic block strategy is greatly reduced.The proposed method in this thesis has important significance for improving the performance of the reversible data hiding schemes under medium and small embedded capacity.