A high quality interpolation-based reversible data hiding technique using dual images

Increasing data hiding capacity and reducing cover image distortions are the main objectives of any data hiding technique. Moreover, some applications require the reversibility of the data hiding technique so that the original cover image is exactly recovered in the extraction step. Interpolation-based data hiding techniques have the advantage of providing high data hiding capacity. However, they suffer two drawbacks: they are not truly reversible and introduce high distortions to the cover image. This paper presents a new interpolation-based data hiding technique that is adaptive, truly reversible, vastly reduces the cover image distortion, and takes the sensitivity of the Human Visual System (HVS) into consideration. Unlike other interpolation techniques, our proposed technique eliminates the down-scaling and expansion steps in typical interpolation-based techniques. Instead, it embeds data into the original cover image. It uses a simple, efficient interpolation algorithm to take the sensitivity of the HVS into account by limiting the distortions in smooth regions of the cover image where the HVS is more sensitive to distortions. Using dual cover images and an improved interpolation algorithm achieves reversibility, vastly reduces cover image distortion, and achieves high data hiding capacity. The downscaling and expansion step in typical interpolation-based data hiding techniques results in poor quality cover images with a peak signal-to-noise ratio (PSNR) in the neighborhood of 25 dB. The proposed technique eliminates this step and produces high-quality stego images with 42dBs minimum average PSNR values.