Three-dimensional Prediction-Error Histograms Based Reversible Data Hiding Algorithm for Color Images

Reversible watermarking technologies have rapidly developed in recent years, but there are few research achievements in the area of reversible watermarking technology for color images. Thus, in this study, a reversible data hiding (RDH) algorithm for color images was constructed based on three-dimensional (3D) histogram shifting. The algorithm improves the peak signal-to-noise ratio and reduces distortion by using rhombus prediction to predict the pixel values of each of the RGB channels of a color image to establish 3D prediction-error groups, and by modifying the mappings of the prediction-error groups in the prediction-error histogram such that mappings with high distortion are rejected and new mappings can be constructed. The experimental results proved that, under the condition of the same embedding capacity, the proposed color image-purposed algorithm, which takes advantage of the correlations between RGB channels, can yield lower distortion than conventional RDH algorithms for color images.