An Invisible Hybrid 3D Video Watermarking Robust Against Malicious Attacks

Abstract

Digital watermarking is known as the most suitable solution to protect 3D videos. Hence, proposing a watermarking technique robust against malicious attacks and especially against collusion becomes a crucial challenge for many researchers. In fact, collusion is usually not considered in the assessment of anaglyph 3D video watermarking schemes. In this paper, an invisible and robust watermarking method dedicated to anaglyph 3D videos is proposed. It embeds a signature in the selected mosaics generated from different groups of frames forming the original video. The choice of the mosaic as an embedding target guarantees robustness against collusion which considered as the most dangerous attack for any video watermarking technique. First, several mosaic images are generated from every set of 25 frames of the original video and they are compared using a similarity measure. In order to obtain a high level of invisibility, the signature will not be embedded in all mosaics but only in the ones which are not similar. Second, the signature is inserted on each selected mosaic image using a hybrid embedding scheme based on the least significant bit and the discrete wavelet transform. Finally, all the marked frames are restored from the marked mosaic images and they are recombined to the unmarked ones to reconstruct the marked anaglyph video. To prove the robustness of the suggested method, different manipulations are applied to various anaglyph videos and several metrics are calculated. Obtained results confirm its robustness against the image processing attacks, the enhancement attacks, and the frame-based attacks, in addition to dangerous video attacks such as MPEG-4 and H264-AVC compression and collusion attacks.