Research On Reversible Watermarking Algorithm With High Capacity Based On Histogram Shifting

The research of high-capacity reversible watermarking algorithm based onhistogram shifting Digital watermarking technology has been widely studied as a newdirection in the field of information security recent years. The traditional digitalwatermark embedding process will usually bring embedding distortion that cannot beerased to the original carrier. Although the distortion is usually very small, the degree ofdistortion is still not acceptable for high-fidelity image applications like medical ormilitary field. In order to solve the problem, some scholars put forward the reversibledigital watermarking technology. Its main thought is to embed data or add watermark inthe original information carrier, and recover the undistorted original carrier information.Its main advantage is that the user can extract the embedded useful information from theembedded data or adding watermark carrier signal at the decoding side, and the user canalso resume the original carrier signal without distortion.Based on a lot of analysis of existing research results, the problem of lowsingle-embedding rate in reversible digital watermarking techniques was in-depthstudied and discussed in this paper. The main innovation of this paper is as follows:The tradition difference expansion method can only embed1bit watermarkinformation in the original pixels. In order to embed multi-bit watermark information insingle process to the original pixels, we promoted and improved the traditionalhistogram shifting method. The algorithm embeds1.0bit or more bit watermark in onepixel in terms of the context of image, which makes the single embedding capacityreach2bpp(bits per pixel).Compared to multi-pass embedding, the reversible watermarking algorithmproposed by this paper avoid using pixels which have large prediction error under largeembedding capacity, thereby reducing the maximum amount of modification of thepixel values, reduces the embedding distortion.The adaptive embedding method proposed by this paper takes into fullconsideration the local feature of host image, so that the location of the embeddedwatermark, embedding strength, embedded information adaptively change with thelocal feature. The method that embeds the same bit watermark information in uniform inall original image pixel values is not optimal strategy. To ensure the image quality afterembedding large capacity watermark information, the paper choose the pixels thatmodify little image after embedding data. In the selection of preferred embedded pixelpair, we choose the pixels with small difference, which the difference after expanding will not be too big, the prediction error histogram will be sharper, scalable predictiveerror increase in histogram, movable pixels reduced, so the distortion is small and theimage quality is higher.The experimental results show that: our improved algorithm achieve higherembedding capacity than the traditional method based on integer transform andhistogram shifting. Under the same capacity, the improved algorithm has smallerembedding distortion compared with multi-pass embedding.