Reversible Data Hiding With Pattern Adaptive Prediction.

The article focuses on improving reversible data hiding (RDH) based on prediction-error expansion (PEE) by introducing pattern adaptive prediction (PAP) using local binary patterns (LBP) derived from a pixel’s eight-neighborhood. PAP clusters pixels according to LBP groups to create multiple prediction-error histograms (PEHs), which, combined with local complexity (LC) measures, form two-dimensional PEHs (2D-PEHs) for optimized data embedding via multiple histograms modification (MHM). To reduce computational complexity, LBP patterns are grouped based on histogram concentration characteristics, enabling efficient performance optimization. Experimental results on standard test images demonstrate that PAP improves prediction accuracy, leading to better image fidelity and embedding capacity compared to existing predictors, with manageable computational overhead.