High-capacity reversible data hiding in encrypted images based on difference image transfiguration

Reversible data hiding in encrypted images (RDHEI) has garnered significant attention from researchers over the last decade, mainly due to its potential to embed secret information into encrypted images without referencing the original image. However, most state-of-the-art (SOTA) RDHEI methods do not provide optimal embedding capacity (EC) and full reversibility of the image. To bridge this gap, a novel RDHEI method based on difference image transfiguration is proposed. The method initially employs the Median edge detector to predict pixels of the image, subsequently deriving a difference image. Some of the most significant bit planes of the difference image are then cleaned in order to create a large space for embedding the secret information. Additionally, this cleaning process results into a transfigured image that enhances the visual security. However, at the same time, it also generates a weight matrix which is essentially nothing but an overhead, that must be recorded for reconstruction of the original image. To significantly reduce the overhead size, the weight matrix is first encoded using a novel location difference technique to compressible lists, and then Huffman encoding is employed. Subsequently, the transfigured image is encrypted along with the compressed overhead and the secret information is embedded in the reserved space. As a result, the method ensures high EC while preserving the complete reversibility of the image. Comparative experiments validate the efficacy of the proposed method relative to the SOTA methods.