High-Dimensional Histogram Utilization for Reversible Data Hiding

Histogram-shifting (HS) is an efficient technique for reversible data hiding (RDH). In this paper, in contrast to previous HS-based RDH works that utilize one or two dimensional histogram, we propose a new RDH scheme exploring three-dimensional histogram. Referred to the recently proposed general framework of HS-based RDH, we first divide the host image into non-overlapping pixel triples to generate a three-dimensional histogram, and then utilize each triple for embedding or shifting based on an applicable triple classification. To further exploit the image redundancy to enhance the embedding performance, we apply the pixel-selection technique in our scheme such that smooth pixel triples are priorly embedded. Experimental results have verified the superiority of the proposed scheme over some state-of-the-art RDH works.