Adaptive Multiple Peak and Histogram based Reversible Data Hiding

Abstract

The aim of this study is to propose a histogram shifting-based reversible data hiding scheme for anatomical magnetic resonance images (MRI) of human brain. In this study, a human brain MRI dataset was employed for experimental purposes, which were obtained from the Alzheimer's Disease Neuroimaging Initiative (ADNI) open dataset. In experimental design, only healthy subjects are included in order to avoid any bias due to morphological brain abnormalities. The dataset contained three-dimensional 16-bit-depth T1-weighted brain MRIs of 25 healthy people. The proposed data hiding scheme for an MRI volume started with an atlas-based partitioning procedure to segment the volume into several region-of- interests, called as chunks in this paper. Then, the well-known histogram-shifting method was applied with some modifications to each chunk, separately. Finally, the total capacity of chunks in bit-per-voxel (BPV) and the stego-image quality in peak signal-to-noise ratio (PSNR) were calculated. The average PSNR and BPV values of the cohort were reported in several experimental conditions including different segmentation regions and several numbers of peak-zero bin-pairs per chunk. The main contribution of this study is the modifications on the traditional histogram-shifting-based reversible data hiding scheme that are considering the characteristics of the T1- weighted MRI data. Furthermore, the distinguishing feature of this study is that, on contrary to similar studies in the literature, this study employed three-dimensional MRI scans of real cohort. The experimental results of this study revealed the applicability of data hiding schemas on neuroimaging data. © 2019 IEEE.