Non-Integer Expansion Embedding for Prediction-Based Reversible Watermarking

Abstract

This paper aims at reducing the embedding distortion by improving predictor’s performance for prediction-error expansion (PE) based reversible watermarking. In the existing PE embedding methods, the predicted values or their variety should be rounded to integer values. This will restrict predictor’s performance since the prediction context is only with past pixels (image) or samples (audio). In this paper, we propose a non-integer PE (NIPE) embedding approach, which can proceed non-integer prediction errors for data embedding by only expanding integer element of a prediction error while keeping its fractional element unchanged. More importantly, the NIPE scheme allows the predictor to estimate the current pixel/sample not restricted only past pixels/samples. We also propose a novel noncausal prediction strategy by combining past and future pixels/samples as the context. Experimental results for some standard test clips show that the non-integer output of predictor provides higher prediction performance, and the proposed NIPE scheme with the new predicting strategy can reduce the embedding distortion for the same payload.