Abstract: This paper proposes a color image blind watermarking algorithm based on Chebyshev neural network. First of all, the algorithm divides the color image into blocks; secondly, it selects the appropriate training samples to train the neural network, and sets up the network model to extract the watermarks; finally, uses the algorithm based on neural network to embed and extract the watermark in the color image. The experiments’ results show that this watermarking algorithm has better robustness and invisibility. Meanwhile this paper also proposes a new watermarking similarity standard with the nature of Clifford vector dot product and cross product, the experiments show that the evaluation standard can detect the similarity of the original watermark and the similarity of watermark.

Key words: network model, Clifford vector, evaluation standard