A Scalable Reversible Data Embedding Method with progressive quality degradation functionality

This paper proposes a novel reversible information hiding method aiming to achieve scalable carrier capacity while progressively distorting the image quality. Unlike the conventional methods, the proposed method HAM (Histogram Association Mapping) purposely degrades the perceptual quality of the input image through data embedding. To the best of our knowledge, there is no method that attempts to significantly increase the carrier capacity while introducing (tolerating) intentional perceptual degradation for avoiding unauthorized viewing. HAM eliminates the expensive pre-processing step(s) required by the conventional histogram shifting data embedding approach and improves its carrier capacity. In particular, the host image is divided into non-overlapping blocks and each block is classified into two classes. Each class undergoes different HAM process to embed the external data while distorting quality of the image to the desired level. Experiments were conducted to measure the performances of the proposed method by using standard test images and CalTech 101 dataset. In the best case scenario, an average of ~2.88 bits per pixel is achieved as the effective carrier capacity for the CalTech 101 dataset. The proposed method is also compared with the conventional methods in terms of carrier capacity and scalability in perceptual quality degradation.