Optimizing Pixel Predictors Based on Self-Similarities for Reversible Data Hiding

This paper presents a clustering and optimizing pixel prediction method for reversible data hiding, which exploits self-similarities and group structural information of image patches. Pixel predictors plays an important role for current prediction-error expansion (PEE) based reversible data hiding schemes. Instead of using a fixed or a content-adaptive predictor for each pixel independently, we first employ pixel clustering according to the structural similarities of image patches, and then for all the pixels assigned to each cluster, an optimized pixel predictor is estimated from the group context. Experimental results demonstrate that the proposed method outperforms state-of-art counterparts such as the simple rhombus neighborhood, the median edge detector, and the gradient-adjusted predictor et al.