Hybrid reversible watermarking algorithm using histogram shifting and pairwise prediction error expansion

Abstract

The exact retrieval of both i.e., watermark and host data using reversible watermarking makes it more suitable for applications like medical, military etc. The techniques which are considered most efficient and are being most widely used to implement it (reversible watermarking) are Prediction Error Expansion and Histogram Shifting as they offer higher embedding capacity with lesser distortion. The methodology proposed in this work is a hybrid one that utilizes the best of both. The algorithm starts with splitting the image into two regions and further processing is done on them one by one thereby covering each pixel for embedding thus amounting to enhanced capacity. The proposed work exploits the histogram of prediction errors instead of difference of adjacent pixels as it is more sharply distributed. This leads to lesser distortion in watermarked image. In this paper, the sorting of prediction errors is done in accordance with the calculated variance values of their prediction context followed by pairwise payload embedding. The results have substantiated the fact that for the smaller variance, prediction errors are also small. The proposed methodology is tested against the standard test images and real life applications such as medical, biomedical, aerial, military and colour images of Kodak image dataset. The experimental results demonstrate the superiority of the proposed work over the existing ones in terms of embedding capacity and distortion.