An Improved Reversible Data Hiding Using Pixel Value Ordering and Context Pixel-Based Block Selection

Abstract

In recent years, pixel value ordering (PVO) and prediction-error expansion (PEE) methods are combined to better exploit the pixel correlation in reversible data hiding (RDH) techniques. In these techniques, the cover image is split into non-overlapping blocks of equal size. The pixels of each block are sorted in ascending, according to their intensity values. The maximum and minimum pixels within each block are used to calculate prediction errors which are either expanded or shifted for embedding data. The data are embedded into smooth blocks which are identified using every two adjacent pixels of neighbor block. However, this block classification is not very effective since it considers only the neighbor block and ignores the pixels in current block. In this paper, to overcome this drawback and to improve the performance further, a novel reversible data hiding using context pixel-based block selection (CPBS) strategy is proposed. After the block division and sorting process, the complexity of each block is measured using the context pixel chosen from the current block and the connected pixels of neighbor block. Smooth blocks with low complexity are determined using a threshold value to achieve higher efficiency. When the desired capacity is achieved, the index orders of all pixels in the block are preserved and thus the reversibility is guaranteed. Experimental results verify that the proposed scheme achieves better capacity-distortion trade-off and thus outperforms with the existing PVO-based techniques.