Effective reversible data hiding scheme for interpolated images using an improved data encoding strategy

Abstract

In 2021, Lu et al. proposed a reversible data hiding (RDH) scheme for interpolated images based on a multilayer center-folding strategy with a high embedding rate and image quality. However, their data encoding method and the chosen block size are not optimal, leaving the watermarked image quality and the embedding rate to be improved. To further reduce the distortion caused by embedded data, we propose an improved RDH scheme for interpolated images using an improved data encoding method. First, an improved two-layer data encoding method is proposed, which encodes the original data ${\varvec{R}}$ by comparing the amplitude of ${\varvec{R}}$ and its complement and the center folding strategy to form a new encoding pair $({\varvec{R}}^{''}, {\varvec{F}})$. Thereafter, the neighbor mean interpolation method proposed by Jung and Yoo is used to generate an interpolated image divided into ${\varvec{m}}\times {\varvec{m}}$ non-overlapping blocks. Additionally, the interpolated pixels in each block are pre-processed to solve the overflow/underflow problem. Finally, the encoded data ${\varvec{R}}^{''}$ and ${\varvec{F}}$ are embedded into the pre-processed interpolated pixels to achieve blind extraction. We theoretically prove that the image quality obtained with a block size of $2\times 2$ is superior to that of many other blocks. The results of extensive experiments on the BossBase, BOWS-2, NCID, and SIPI databases confirm that the proposed scheme outperforms several state-of-the-art schemes, with the average PSNR value improved by $\sim$ 7.58 dB when 3-bit data are embedded into each interpolated pixel. These results indicate that our proposed scheme is effective.