Abstract: We propose a nonlinear,adaptive,gradient-adjusted predictor based on full context (FCGAP)．FCGAP is suitable for reversible image data hiding．In FCGAP,the predicted value of target pixel is equal to the weighted average of 4neighborhood pixels,and the weighted coefficients are computed by 12nearest neighbor pixels．Based on FCGAP,a reversible data hiding is proposed,which can provide high data embedding capacity and low distortion．Experimental results demonstrate that FCGAP can more fully utilize context information,and has better prediction accuracy of pixels than other predictors；the reversible data hiding based on FCGAP has better capacity-distortion performance compared with other state-of-the-art algorithms.

Key words: Reversible data hiding,Lossless data hiding,Reversible image watermarking,Prediction error expansion,Full context prediction