Reversible Data Hiding Using Improved Gradient Based Prediction and Adaptive Histogram Bin Shifting

Abstract

In this paper, a novel reversible data hiding method is proposed using an improved gradient based prediction. The gradients are computed across horizontal, vertical, diagonal and anti-diagonal directions using 5 × 5 neighborhood. The final predicted value is obtained as a weighted average of two linearly predicted values in the directions of the smallest two gradients. An adaptive prediction error histogram bin shifting method is also proposed to insert either 1 bit or 2 bits in the prediction error of a pixel. It is experimentally observed that the proposed reversible data hiding technique using an improved gradient based prediction and adaptive prediction error histogram bin shifting outperforms several other existing techniques.