Efficient Reversible Data Hiding Based on View Synthesis Prediction for Multiview Depth Maps

Abstract

Multiview-plus-depth (MVD), which has been used as a 3D format, consists of a texture image and its corresponding depth map at each viewpoint. This MVD-based 3D format has two strong correlations, which are an inter-component correlation between a texture image and its corresponding depth map and an interview correlation between adjacent texture and depth views. However, conventional reversible data hiding (RDH) methods have been mainly developed for texture images. As a result, high performance could not be achieved in depth maps. In order to solve this problem, an efficient RDH method is proposed for multiview depth maps in this paper. The proposed RDH method checks inter-component and interview correlations, and uses inter-component and interview predictions adaptively to embed secret data in the depth map. In particular, a view synthesis prediction (VSP) method is used for the interview prediction. In addition, an allowable depth distortion range, which guarantees no synthesis distortion in virtual views, is calculated to minimize distortion of the depth map marked with the hidden data, while maintaining the high embedding capacity. Experimental results show that the proposed method achieves much higher performance than conventional methods in terms of the embedding capacity and the depth distortion.

Keywords

• Depth map
• multiview-plus-depth (MVD)
• reversible data hiding (RDH)
• view synthesis prediction (VSP)