A Reversible Data Hiding Using Difference-Histogram Modification on Multi-directional in Two-Dimensional Histogram

摘要

Reversible Data Hiding (RDH) also called as lossless compression is a technique which is used to recover Cover Image (CI) and the embedded data from Watermarked Image (WI) without any loss of data. Cover image data get altered due to embedded data. Retrieving cover image and embedded data from watermarked image, without any loss, is a challenging task. To solve that problem using RDH, RDH introduced a new trend in fields such as telemedicine, military, and law enforcement, where the original pixel should be recovered fully. Especially in medical images, getting error in recovered data is intolerable which may lead to miss diagnosis [1]. Earlier, different methods were proposed in RDH, such as Jun Tian introduced a new scheme in RDH using Difference Expansion (DE). Difference expansion embeds the data with the help of 1D Haar wavelet transform [2]. Different approaches have been proposed for RDH using histogram modification [3-5], prediction error expansion [6-8], and integer transform [9-11]. Among all these approaches, histogram modification got more popular in the field of RDH. The first proposal started with Zhicheng et al. [5]. Zhicheng et al. embedded data by using peak (P) and min (Z) pole from a histogram. This method fails to handle images that contain equal height poles in a histogram. A separate channel is required to send "P", "Z" values for successful restoration of embedded data and cover image. Tai et al. [3] introduced RDH using pixel difference. This method avoids the drawbacks in Zhicheng et al. Author used Binary Tree Structure (BTS) to introduce more redundancy in multiple levels. The height of the tree (L) is used for successful restoration of CI and embedded data. Hong [12] introduced RDH using Double Binary Tree (DBT) to increase more Embedding Capacity (EC) and PSNR compared with Tai et al.