A robust rotation resilient video watermarking scheme based on the SIFT

This paper proposes a blind and robust video watermarking algorithm for copyright protection of the video in the wavelet transform. The watermark bit is embedded by modifying two random high frequency coefficients in the first level discrete wavelet transform of the frames of the video. The watermark is encrypted before embedding by using the cat map for better security. During the watermark extraction phase, it is verified whether the video is attacked by rotation attack. Frame restoration against rotation attack is done based on matching at least two feature points each generated by the scale invariant feature transform in the original video frame and the watermarked one. The angle between the two lines formed by joining these feature points is used to restore the video frames to the original position. Experimental results show the imperceptibility and robustness of the proposed method against variety of attacks. It is especially robust against rotation attack in comparison with other popular watermarking methods. Normalized Correlation, Structural Similarity Index and Peak Signal-to-Noise-Ratio are nearly 100%, 1, above 48 dB respectively against the various attacks.