Research On Reversible Watermarking Algorithm Based On Mean Adjustment And Total Difference Expansion

Reversible watermarking is a special kind of watermarkings that the cover image can be completely recovered after extracting the embedded data. By Compared with traditional watermarkings, it has strict requirements on the cover image recovery, and commonly used for protection of important images without distortion such as military, medical and remote-sens ing images.Currently, many kinds of reversible watermarking algorithms have been proposed, such as difference expansion based reversible watermarking algorithm, histogram shifting based watermarking algorithm, prediction-error expansion based reversible watermarking algorithm, mean-adjustable integer transform based reversible watermarking algorithms, and local complexity and generalized prediction-error expansion based reversible watermarking algorithm. In these watermarking algorithms, mean-adjustable integer transform based reversible watermarking algorithm and local complexity and generalized prediction-error expansion based reversible watermarking algorithm can achieve high embedding capacity, but they still have some defects. In the existing mean-adjustable integer transform based reversible watermarking algorithms, there are following defects such as non-adaptive threshold selecting, incomplete location map building strategy which may lead to poor compression performance and flawed embedded vector partition strategy embedding algorithm which may lead to a failure embedding even if embedding capacity is enough, In the existing local complexity and generalized prediction-error expansion based reversible watermarking algorithms, there are following defects such as the maximum embedding capacity depends on the payload data, threshold can not guarantee maximum capacity, the normalization in pixel classification reduces the accuracy, strategy that recording all overflow pixels'locations have a great effect on the maximum embedding capacity, generate and embed backup data in cover image after embedded payload data directly may cause the watermarking algorithm irreversible and the high embedding parameters selection algorithm's time complexity.To address these problems, the followings are the finished works:1)In the existing mean-adjustable integer transform based reversible watermarking algorithms, there are following defects such as non-adaptive threshold selecting, incomplete location map building strategy which may lead to poor compression performance and flawed embedded vector partition strategy embedding algorithm which may lead to a failure embedding even if embedding capacity is enough. To address these problems, an iterative adaptive reversible image watermarking algorithm combined with mean-adjustable integer transform was proposed. In the proposed method, firstly an iterative adaptive algorithm was used in selecting mean-adjustable offsets according to Peak Signal-to-Noise Ratio (PSNR) which balanced embedding capacity and visual quality; Secondly based on strategy that pixels adjacent have similar pixels values, a complete location map generating strategy was proposed to improve location map compression performance; Finally to avoid failure embedding, the proposed reversible watermarking algorithm adopted hierarchical order embedding strategy to embed payload data in order from the least significant bits to the third least significant bits. Experiments show the proposed algorithm has a big embedding capacity and no need to preset threshold. Location embed map building strategy has a better performance in making location map data in smaller size and increasing the capacity indirect by comparing with mean-adjustable integer transform based reversible watermarking algorithms. The average increase of PSNR is 14.4% in samples by compare with mean-adjustable integer transform based reversible watermarking algorithms.2) In the existing local complexity and generalized prediction-error expansion based reversible watermarking algorithms, there are following defects such as the maximum embedding capacity depends on the payload data, threshold can not guarantee maximum capacity, the normalization in pixel classification reduces the accuracy, strategy that recording all overflow pixels' locations have a great effect on the maximum embedding capacity, generate and embed backup data in cover image after embedded payload data directly may cause the watermarking algorithm irreversible and the high embedding parameters selection algorithm's time complexity. To address these problems, a local complexity and direction changeable prediction-error expansion based watermarking algorithm combined with sort and enumeration was proposed. In the proposed method, firstly direction changeable prediction-error expansion was used to increase the maximum embedding capacity and makes the maximum embedding capacity is independent of payload data at the same time, and set different thresholds for different pixels' types makes possible for maximum embedding capacity; Secondly the proposed method avoided normalization in pixel classification and improved the match accuracy; Thirdly the single layer location map was used to records all overflow pixels reduced the relevant effect on the maximum embedding capacity brought by overflow pixel processing; Fourthly the additional data and backup data are synchronized embedded into the cover image to guarantee watermarking algorithm completely reversible; Finally embedding parameters selection strategy based on sorting and enumeration has a lower time complexity. Experiments show that, by compared with the existing method, the proposed algorithm has a larger maximum embedding capacity and it is completely reversible, the desired time of embedding parameters selection strategy is decreased significantly.