Gradient dependent reversible watermarking with low embedding impact
Ravi Uyyala, Rajarshi Pal, Munaga V. N. K. Prasad · 2016
This paper presents a novel prediction error estimation (PEE) based reversible watermarking scheme. A good predictor is key to the performance of this kind of watermarking scheme. A novel gradient based predictor estimates the pixel value based on a 5×5 neighborhood of the pixel. Moreover, the prediction error expansion (PEE) is divided between the current pixel and its top-diagonal neighbor such that distortion remains minimum. Experimental results establish that the proposed predictor with optimal embedding outperforms several existing methods.