Adaptive reversible watermarking using trimmed prediction and pixel-selection-based sorting

Prediction error expansion based on sorting is an important technique in reversible watermarking since it yields large embedding capacity and low distortion. In this paper, an efficient and adaptive reversible watermarking scheme is proposed based on trimmed prediction and pixel selection sorting. The trimmed prediction excludes one singular pixel from the neighboring region. A more efficient sorting method is used to achieve lower distortion. Then, a further sorting that considers context complexity is proposed to ensure better visual quality. The smooth pixels located in rough areas are assigned high priorities for carrying bits by using the prediction error expansion method. With these improvements, our method shows better performances in terms of capacity and distortion.