Reversible 3D Image Data Hiding with Quality Enhancement

摘要

The 3D image constructed by a texture image plus a depth map is popular for stereo representation. When embedding data in the depth map, conventional reversible data hiding algorithms, which are designed for gray and color images, are not suitable. In this paper, we propose a novel reversible data hiding method in the depth map for quality enhancement. By optimizing the selection of prediction error and embedding data with the help of depth no-synthesis-error model (D-NOSE), the proposed scheme can embed more data with less distortion. The hidden data are the compressed residual error between the actual view and the virtual view. The watermarked depth map can directly be used to synthesis a same virtual view as in depth-image-based rendering (DIBR). After extracting the hidden data, a higher quality virtual view can be obtained by adding the residual error to the synthesized virtual view. Experimental results show that the virtual view with higher quality is obtained, and higher embedding capacity is achieved as well.