Contourlet transform based digital watermarking resisting 2D-3D conversion

Abstract

Nowadays with the success of 2D-3D applications, the copyright protection for such cases is becoming increasing important. In this paper, a contourlet transform based watermarking scheme is proposed, which is robust against DlBR-based 2D-3D conver sion. In the proposed scheme, the contourlet transform is per formed on the center image. Considering that the DlBR trans form can be regarded as the horizontal direction movement, coefficients of the low-frequency after contourlet transform are quantized to embed the watermark. Experimental results show that the watermark can be extracted from the virtual left and right image correctly and stably. The proposed scheme achieves better robustness and invisibility, which demonstrates that it is promising in 2D-3D conversion applications. Index Terms -Digital Watermarking; Copyright Protec tion; DlBR; 2D-3D Conversion; Contourlet Transform 1. INTRODUCTION Recently, different kinds of 2D-3D conversion applications are arising with the emerging of low cost 3D display devices. The digital watermark for such 3D media protection must play an important role for developing and promoting the 3D entertain ment industry. According to [1], the 3D watermarking systems can be classified on the basis of their embedding space and detecting space as: 3D/3D, 3D/2D and 2D/2D. The 3D/3D watermarking system embeds the watermark into a 3D (geometry or texture) model and detects the watermark also from the 3D space, while the 3D/2D watermarking system extracts the watermark in a 2D representation (acquired by projection) of a 3D data. The 2D/2D watermarking deals with the content protection of image-based 3D data representation in which all the embedding and detection operations are performed on 2D images directly. In 2D-3D conversion, a central image and the correspond ing depth image are usually needed. The related two virtual im ages corresponding to the views in human eyes are generated by the central image and associated per-pixel depth information with the DIBR technique [2]. So these two virtual images are similar to the central image perceptually. These two virtual images can be distributed as 2D images again and it causes a new copyright protection challenges. As is known to all, the copy right protection should be robust to DIBR process first and so the traditional digital watermark algorithms for 2D images can't