A watermarking algorithm for image authentication was designed

which was based on second generation Bandelet translation and statistical model.The image directional flow was constructed by using Bandelet

and it could depict the characteristic of the image.By applying noise visibility function

the noise visibility intensity algorithm which was based on Gaussian statistical models was proposed.The blocks of image which contain more texture information were selected to hide the image directional flow

based on the noise visibility intensity algorithm.The image was authen-ticated through comparing the hiding information and the carrier image directional flow created by Bandelet.Experimen-tal results show that the algorithm can resist the JPEG compression and noise attack strongly and guarantee the quality of the sight.It is a blind detection algorithm

and this algorithm has high concealment performance and security.