Research On Robustness Of Reversible Image Watermarking

As a new watermarking technique, robust reversible image watermarking attracts more attention recently years. Robust reversible watermarking, that is, the watermark can be extracted from watermarking image and original image can be correctly restored when watermarking image does not suffer from malicious attacking or conventional editing. The watermark can still be correctly extracted under certain attack. This technique, highly robust and reversible, has been widely used in steganography for highly sensitive and secret images.By studying previous robust reversible image watermarking, we proposed two improved reversible watermarking algorithms:1. A novel robust reversible watermarking based on image dichotomy. The original image is firstly divided into two sub-images of the same size, the watermark can be embedded into one of these sub-images with a robust watermarking algorithm based on DCT to ensure highly robust. Because the DCT is not a kind of integer transform, there will be residual error between the original image and the recovered image after the watermark is extracted, and the residual error should be embedded into the other image to ensure the reversibility in the algorithm with discrete histogram shifting, an algorithm of highly embedding rate. Finally the two sub-images is combined into one image. This algorithm is highly robust and reversible.2. A robust reversible image watermarking based on time domain JND. With this algorithm, the original image can be divided into two regions of texture region and smooth region. Firstly the size and the location can be determined by embedding capacity. After that the watermark is embedded into the texture region based on DCT image watermark algorithm of mid frequency to ensure highly robust and make a smaller difference in visibility. A robust reversible image watermarking based on time domain JND. With this technology, the original image can be divided into two regions of texture region and smooth region. Firstly the size and the location can be determined by embedding capacity. After that the watermark is embedded into the texture region based on DCT image watermark algorithm of mid frequency to ensure highly robust and make a smaller difference in visibility. The residual error should be embedded into the smooth region to ensure the reversibility in the algorithm with discrete histogram shifting, an algorithm of highly embedding rate. This algorithm not only has the advantages of high robustness and reversibility, but also obtains the better visibility.In this paper, two algorithms above are both robust and reversible watermarking algorithms for gray image. The simulation experiments show that the first algorithm hasthe high robustness and reversibility, and that the second algorithm, under the condition of the domain JND technology, implements the adaptive selection of embedded blocks and achieves a better performance. Our proposed methods have important implications for robust reversible image watermarking.