Study On Histogram Shifting Based Reversible Image Watermarking

As a new watermarking technique, reversible image watermarking attracts more attention in recently years. On the basis of watermarking extracted correctly, reversible image watermarking should remove image distortion exactly, restore cover image. Since the reversible watermarking method transmits information without destroying the cover image, this technique has been widely used in steganography for highly sensitive and secret images.By studying previous reversible image watermarking, we proposed two improved reversible watermarking algorithms:1. A novel reversible watermarking based on dynamic blocking and pixel-value-order. PVO is a kind of predicting technique, by sorting pixels in block in ascending order, maximum(minimum) pixel in the block can be predicted by the second large pixel(the second small pixel). By calculating difference between predicting pixel and original pixel, a difference histogram can be made. Watermarking bits prior embed in smooth block to achieve a sharper histogram. Dynamic blocking uses multi-thresholds to control block size according to noise level, that resolve the problem that fixed size block cannot adjust flexible by embedding capacity.2. A reversible image watermarking based on discrete histogram shifting. We proposed a reversible image watermarking algorithm by taking use of discrete histogram bins to expanding and embedding. Firstly, GAP is used to predicting, after calculating difference between predicting and original pixel, we get difference histogram. For each embedding capacity, we choose at most four bins to expanding, and other bins is shifted to provide reversibility. While traditional histogram shifting uses continuous bins for extension and other bins are shifted. Because expanding provides embedding capacity, shifting only introduces distortion, our proposed method adaptively chooses bins to expanding according to the EC, that decrease unnecessary shifting. Therefore the proposed method achieves a better performance.In this paper, the two reversible watermarking algorithms we proposed are suitable for low EC. Experiences show that the first algorithm improves the performance on larger EC, while the second algorithm improves the performance when multi-bins are used. Our proposed methods realize dynamic blocking and bins adaptive selecting respectively according to EC and image content and achieve a better performance. Our proposed methods have important implications for reversible image watermarking on low EC.