Dual-image-based reversible data hiding method using center folding strategy

In recent years, dual-image techniques have been widely used in reversible data hiding. Most data hiding methods focus on finding the best predictive value to reduce the degree of stego image distortion. After finding a predicted value, most methods convert the secret data directly into secret symbols, using the decimal or other numeral systems, and then have these added to (or subtracted from) the predictive value (or pixel) to complete the embedding. However, this study found that the secret data are a very important factor affecting image quality. If the value of the secret symbols is too large, it may cause greater differences between the stego-pixels and original pixels. Hence, this study employs the center folding strategy to reduce the value of the secret symbols. The reduced symbols are then embedded in two stego images through an averaging method, to maintain image quality. In addition, underflow/overflow pixels are replaced with the original pixels to reduce unnecessary image distortion. The experimental results show that the proposed method has fairly good performance compared to the other dual-image techniques. For the same embedding capacity, its image quality is also better, on average, than that of other methods by at least 2dB. Furthermore, the proposed method determines the embedding capacity with the k value. The larger the k value, the higher the embedding capacity. The experimental results show that for the same image quality, the embedding capacity of the proposed method is better, on average, than that of other methods by at least 0.5 bits per pixel. The proposed method is applicable not only to experimental images, but also works well for general images in terms of embedding capacity, image quality and execution time. This research presented a dual Imaging-based reversible hiding technique.The Central Folding Strategy is applied to achieve low distortion purpose.The reduced symbols are embedded in two stego-images through an averaging method.The performance of the proposed method is better than that of other methods.The proposed method can adapt to different image characteristics.