The reversible data hiding is an emerging technology that uses the redundancy of the carrier (typically digital images) to embed secret information and ensure the reversibility of the carrier and hidden information. In recent year, a number of reversible data hiding algorithms based on prediction error expansion have been developed. In prediction error expansion, prediction on the center pixel is made based on its neighbor pixels. The data embedding is conducted by the modification on the histogram made from prediction error expansion. Therefore, the accuracy of prediction on pixel is the key to improve the performance of the algorithm. In this paper, we propose a new reversible data hiding based on directional prediction and multiple histograms modification and design the corresponding reversible hiding rules. Compared to the existing algorithms, experimental results show that the proposed method can reduce distortion to the image at given embedding capacity.
Recently, more and more attention is paid to reversible data hiding (RDH) in encrypted images because of its better protection of privacy compared with traditional RDH methods directly operated in original images. In several RDH algorithms, prediction-error expansion (PEE) is proved to be superior to other methods in terms of embedding capacity and distortion of marked image and multiple histograms modification (MHM) can realize adaptive selection of expansion bins which depends on image content in the modification of a sequence of histograms. Therefore, in this paper, we propose an efficient RDH method in encrypted images by combining PEE and MHM, and design corresponding mode of image partition. We first divide the image into three parts: W (for embedding secret data), B (for embedding the least significant bit(LSB) of W) and G (for generating prediction-error histograms). Then, we apply PEE and MHM to embed the LSB of W to reserve space for secret data. Next, we encrypt the image and change the LSB of W to realize the embedding of secret data. In the process of extraction, the reversibility of image and secret data can be guaranteed. The utilization of correlation between neighbor pixels and embedded order decided by the smoothness of pixel in part W contribute to the performance of our method. Compared to the existing algorithms, experimental results show that the proposed method can reduce distortion to the image at given embedding capacity especially at low embedding capacity.