Abstract: In order to realize the image watermarking as a function of data authentication, access control and copyright protection, a wavelet transform digital image watermarking algorithm which gave a full consideration to the human eye visual perception characteristics was proposed. First, Arnold transform and the stacking method were performed on the binary image watermarking, which was converted into the one-dimensional sequence; second, the kind of modulation was adopted when embedding the watermark information determined by the positive and negative values of key generation gaussian sequence; and then, the original image was processed with 3-level discrete wavelet transform(DWT) decomposition, on the second floor of the wavelet image intermediate frequency sub-band, combining with the human visual characteristics, the positive or negative modulation of watermarking was superimposed on wavelet coefficients with the host image greater than the just noticeable difference(JND) value. The experimental results show that, the algorithm can resist JPEG compression, plus-noise, shearing and other attacks, and has good transparency, robustness and security.