Very effective multi-layer reversible embedding method using low distortion modification

In recent years, prediction-based reversible information hiding approaches have been widely developed. For example, Lee and Chen proposed an adjustable prediction- based reversible information hiding approach, hiding capacity of which is higher than that of previous approaches. However, the altered pixels used for embedding secret data are used to predict the next pixel. The prediction results decrease significantly because the previous pixels may suffer from the large modification. In order to solve this problem, in this paper, we propose a new approach with low distortion modification, which modifies each pixel by increasing or decreasing at most one gray level. Thus, the distortions to cover pixels are slight. Moreover, in the proposed approach, some pixels are not altered to enhance the prediction efficiency in the next embedding round. Consequently, the proposed approach is very suitable for multi-layer embedding. Experimental results show that the proposed approach has better image quality than recent prediction-based reversible information hiding approaches under the same hiding rate. © 2015 ISSN.