Reversible Watermarking Based on Histogram Shifting of Difference Image between Original and Predicted images

摘要

Reversible watermarking is a technique that can recover an undistorted original image from a watermarked image. The proposed watermark embedding algorithm uses histogram shifting of the difference image between a modified original image and its predicted one. In the proposed algorithm, the predicted image that works well increases the embedding capacity, so that the reference pixels for prediction are adaptively selected and filtered and the other predicted pixels are directionally interpolated with the reference pixels. The simulation results demonstrate that the proposed algorithm generates good performances in the peak signal-to-noise ratio (PSNR) values and the embedding capacity.