Enhancing the embedding payload by handling the affair of association and mapping of block pixels through prediction errors histogram
A. H. M. Kamal, Mohammad Mahfuzul Islam · 2016
Images are intentionally distorted in many reversible data hiding schemes. This distortion is performed either by encrypting before data embedment or by applying histogram association mapping (HAM) during data concealment. The later one minimizes the processing complexity and demolishes the requirement of sharing secret keys. In the HAM, each block's pixels belong to a segment in the gray scale are associated and mapped by the message bits to another gray part. The quantity of embedded bits are determined by the range of block pixels, i.e. limit within which pixels spread out. Though, smaller valued ranges contribute much to embedding payload, the frequency of such ranges decreases in archetype images. However, a predictor, if applied on the block pixels, huddles its prediction errors together. Consequently, the range of absolute values of the prediction errors become smaller. The proposed scheme, therefore, applies HAM to prediction errors and shifts the block pixels accordingly. Experiments were conducted using three different standard image datasets to examine the enhancement in the embedding payload and to investigate the performance on different block-sizes. The investigation reveals that the average payloads in the proposed scheme is 1.5 to 2 times of the competing one in all the cases.