Reversible Video Watermarking Based on Inter-prediction Error Expansion

To increase the embedding capacity and to improve watermarked video quality of reversible video watermarking,a novel algorithm using motion estimation and modification of prediction error histogram is proposed.On the embedding side,prediction errors of each frame are obtained by motion estimation considering the relationship among adjacent frames of a video;histogram of prediction errors is generated,and the prediction errors at the peak points of the histogram are expanded to hide data.A little head information to be used in the watermark extracting process,including motion vectors,boundary map,etc.,is appended by the watermark and embedded into the reference frame.On the extracting side,the watermark and head information hidden in the reference frame are retrieved and the reference frame is recovered first,which guarantees a correct context for the extraction of the current frame;then the hidden data in the current frame is extracted according to the head information as well as the recovered reference frame.Experimental results have demonstrated that the proposed algorithm can gain a highly concentrated prediction error histogram and provide a much larger embedding capacity and a higher quality of watermarked video than other reversible watermarking algorithms.