Reversible data hiding based on adaptive IPVO and two-segment pairwise PEE

Abstract

Improved pixel value order (IPVO) and pairwise prediction error extension (PEE) strategies can effectively improve the hidden capacity and invisibility of reversible data hiding (RDH) scheme since they can increase the effective prediction error rate (EPER). To further improve EPER, this paper proposes an improved RDH scheme by combining adaptive IPVO (AIPVO) with two-segment pairwise PEE. Different from the fixed predicted value (such as the highest value) for all blocks in the existing IPVO, AIPVO adaptively determine the predicted value of each block according to its own complexity level. To obtain the complexity level of each block, the block complexity level division model is constructed and optimized with EPER as the objective function. AIPVO can effectively improve EPER due to the fact that the pixels with large fluctuate in a block are eliminated by adaptively selecting the predicted value. In the embedding stage, the two-segment pairwise PEE is designed according to the distribution characteristics of the prediction error sequence obtained by AIPVO. Specifically, the prediction error sequence is firstly divided into two segments, and the 2D mappings of the front and back parts use small-magnitude and slightly large-magnitude prediction errors to embed data, respectively. Experimental results show that this paper scheme is better than the latest research results. For example, PSNR of the Lena image reaches 61.55 dB under capacity of 10,000 bit, and the gain is 0.39 dB against the best result in the literature.