A robust watermarking scheme based on mean quantization

A robust watermarking scheme based on mean quantization and human visual system is proposed. Different from conventional quantization methods, each watermark bit is embedded by quantizing the mean value of a coefficient block with size 2times2 so as to make the embedded watermark more robust, and the account of modification is decided according to the characteristics of human visual system to gain better imperceptivity. In addition, we apply image scrambling to the watermark for improving its robustness against image clipping operation and security. The proposed algorithm recovers the watermark without any reference to the original image. Extensive experiments show that the proposed scheme has a good imperceptibility and high robustness to various common image processing attacks.