High Fidelity Reversible Data Hiding For Color Images using CNN Predictor and Reference Error
Sonal Gandhi, Rajeev Kumar · 2024
In the field of reversible data hiding, both prediction and embedding techniques play an important role in defining the overall effectiveness of a data hiding scheme. A lot of efforts have been invested by the research community to improve the embedding performance. However, the prediction area is still evolving, wherein the Convolutional neural network (CNN), with its capability to consider long-range dependencies, has shown a new direction to the researchers. This paper proposes an error adjustment strategy with a CNN-based predictor for color images. The proposed strategy utilizes the inter-channel correlation of the three channels to enhance the prediction performance. Experimental results show that the prediction accuracy achieved by the proposed method is significantly higher than the state-of-the-art methods. Furthermore, using the proposed predictor, the Peak signal-to-noise ratio achieved for the embedded images for most of the standard test images is better than the known state-of-the-art methods.