Watermarking Based Quality Assessment for DIBR 3D Images

The quality assessment of 3D images is playing a critical role in 3D multimedia applications. In this paper, we propose a novel watermarking based quality assessment scheme for depth-image-based rendering (DIBR) 3D images. The scheme utilizes the extracted watermarks to evaluate the quality of the watermarked images under various distortions. In this scheme, the watermark is embedded into the selected dual-tree complex wavelet transform (DT-CWT) coefficients of the center view. The watermark can be detected from the watermarked center view, the synthesized left and right views separately, and the quality of these views under various attacks can be estimated by the extracted watermark degradation. The performance of the proposed scheme is evaluated in terms of both the classical 2D quality metrics and some well-known 3D quality models. The estimated quality and the calculated quality are highly correlated, which demonstrates that the proposed scheme can assess the quality of DIBR 3D images with high accuracy under JPEG compression, JPEG2000 compression, Gaussian noise and Gaussian blur.