Reversible Data Hiding Based on Prediction-Error-Ordering

Jianqiang Qin, Fangjun Huang

Abstract: In this paper, we propose a new reversible data hiding scheme called prediction-error-ordering (PEO). Instead of sorting the pixels directly as that in pixel-value-ordering (PVO) series, the obtained prediction errors in each block are sorted in ascending (or descending) order. After sorting, the maximum/minimum value in each block is predicted with the second largest/smallest value. Since the prediction errors are around zero in general, a steeper histogram with a peak value of 0 can be easily obtained, which means that more elements can be expanded for data hiding in the next embedding process. Combined with a new adaptive embedding strategy, the marked images with better visual quality can be obtained compared with those popular PVO series.