A robust and imperceptible watermarking method for 3D DIBR images

A well-known representation method for 3D images is the Depth-Image-Based-Rendering method (DIBR). It operates on a monoscopic center view and an associated per-pixel depth map image to generate two virtual views; a left-eye view and a right-eye view. During transmission, the three views can be illegally copied and manipulated, and thus providing copyright protection for the different DIBR views is deemed necessary. In this paper, we propose a 3D DIBR image watermarking method that fulfills the robustness and imperceptibility requirements of effective watermarking. The robustness requirement is fulfilled since the proposed algorithm utilizes the properties of the Integer Wavelet Transform using Lifting Scheme (IWTLS) and the Singular Value Decomposition (SVD). The IWTLS-SVD hybrid transform provides the shift invariance property which gives robustness against the translation caused the rendering operation of the DIBR process. Moreover, the Speed Up Robust Features (SURF) technique is used to provide robustness to geometrical attacks. At the sending end, the stereoscopic center image is decomposed into multiple frequency sub-bands by the IWTLS transform, after which selected sub-bands go through SVD transformation. At the receiving end, the watermark bits are blindly extracted from the three DIBR views: the monoscopic center image, the left-eye view, and the right-eye view. Performance evaluation results show that the proposed method is imperceptible, and robust against signal processing operations such as additive noise, JPEG compression, affine transformation, median filtering, re-scaling, rotation, and cropping. Imperceptibility is demonstrated by the relatively high average PSNR and SSIM values: 46.712 and 0.998, respectively, and robustness by the low average bit error rates (BER) values ranging from 0% to 0.4%, for most signal processing and geometrical attacks.