A robust digital watermarking method for depth-image-based rendering 3D video

Abstract

Depth-image-based rendering (DIBR) has become an accessible rendering technology for 3D video. A variety of digital watermarking methods have been proposed to protect the copyright of DIBR 3D video works. However, the robustness and imperceptibility of the existing methods need to be improved. Therefore, we apply the DIBR rendering features to propose a watermarking method to enhance the watermarking effect. First, to improve the robustness, we combine the DIBR rendering rules to construct steady feature data as the selection criterion of watermark embedding position. We detect the scale invariable feature transformation (SIFT) feature points from centre views and match them in every two adjacent views. Each two matched feature points construct one matching vector. The gradient-probability distributions of the vectors are used as the feature data to determine the watermark embedding position. Thus, the embedding positions are robust to the affine transformation and geometric attacks, and the watermark robustness improved. Second, to improve the imperceptibility, we design a joint watermark extraction strategy based on the similarity of the rendered left and right images. The watermark data are embedded into the centre view with low embedding strength and jointly extracted from the left and right images. This strategy guarantees the extraction accuracy while reducing the impact of watermark embedding on the original centre view. Experimental data show that the proposed method has good robustness and imperceptibility.