Concerning the problem that the existence of secret information of the stego-image on wavelet domain is subject to detection by histogram analysis for wavelet coefficient, and the visual property that human eyes are less sensitive to textured areas of an image, a novel steganography on wavelet domain against statistical analysis based on HVS (human visual system) was proposed. First, an image was divided into blocks of prescribed size, and every block was decomposed into wavelet using IWT (integer wavelet transform) based on lifting scheme. Then, all nonzero wavelet coefficients were classified into two categories according to their sign and parity, and marked by steganographic 0 and 1 respectively. If the identifier of wavelet coefficients was different from secret bit, it would be modified to the opposite one, otherwise unchanged. And more secret information was in turn embedded into 2 or more level wavelet coefficients in textured blocks so that good visual imperceptibility was obtained. Finally, the stego-image was obtained by the inverse wavelet transform. From the experimental results, the proposed method effectively keeps the identity of histogram for wavelet coefficients and maintains a good visual quality of stego-image.