A Robust Multi‐Key Reversible Data Hiding Scheme Leveraging Image Duplication and Smoothness Estimation

ABSTRACT Reversible data hiding in encrypted images (RDHEI) has attracted considerable attention for its ability to support secure communication and storage, yet existing multi‐key encryption‐based schemes often suffer from reconstruction inaccuracies under small‐block settings and high embedding rates. Motivated by this limitation, this study aims to achieve highly accurate and fully reversible image recovery even under small‐block and high‐embedding conditions. To address this limitation, we propose an enhanced RDHEI scheme that combines horizontal image duplication with a smoothness‐driven block selection mechanism. The dual‐reference structure effectively enlarges contextual information for pixel correlation, while smoothness‐guided recovery ensures more reliable reconstruction, even in complex textured regions. Extensive experiments on benchmark grayscale images demonstrate that, for a block with 10 embedded bits, the restoration error rate on Lena is reduced from 0.76% to 0.01%, yielding near‐perfect reversibility, infinite PSNR, and SSIM values consistently above 0.98. Compared with conventional approaches, the proposed scheme delivers superior accuracy, robustness, and visual fidelity. These results highlight the method's potential for secure image communication and encrypted image storage, and provide a foundation for future work exploring extensions to multi‐channel (RGB) images, larger datasets, or video sequences.