Cognitive Computation of Compressed Sensing for Watermark Signal Measurement

As an important tool for protecting multimedia contents, scrambling and randomizing of original messages is used in generating digital watermark for satisfying security requirements. Based on the neural perception of high-dimensional data, compressed sensing (CS) is proposed as a new technique in watermarking for improved security and reduced computational complexity. In our proposed methodology, watermark signal is extracted from the CS of the Hadamard measurement matrix. Through construction of the scrambled block Hadamard matrix utilizing a cryptographic key, encrypting the watermark signal in CS domain is achieved without any additional computation required. The extensive experiments have shown that the neural inspired CS mechanism can generate watermark signal of higher security, yet it still maintains a better trade-off between transparency and robustness.