Reversible data hiding for 3D mesh models with hybrid prediction and multilayer strategy

摘要

This paper presents a novel reversible data hiding (RDH) method for 3D mesh models using hybrid prediction scheme and multilayer strategy. In our prediction scheme, the prediction context is determined by the uniformity of mesh region, which means that full context and partial context are used for uniform region and nonuniform region, respectively. This kind of content-dependent prediction can further exploited vertex-vertex correlations to achieve better prediction accuration. Meanwhile, a multilayer strategy with overlapping partition is used to make a trade-off between important requirements: the embedding capacity should be large and distortion should be low. Compared with non-overlapping partition, it keeps not only the prediction context intact but also more capacity available in one pass embedding. In addition, the threshold is adaptively selected considering the magnitude of prediction errors such that the embedding distortion is reduced when embed the secret data into 3D mesh models. Experimental results also validate the effectiveness of the proposed method.