An adaptive algorithm operating in the Contourlet domain is presented. Contourlet is a new image sparse representation, which is better than a wavelet for piecewise smooth images with smooth contours. Because of flexible multiresolution, local and directional sensitivity of Contourlet transform, our approach also defines significant-tree in the Contourlet domain. By analyzing the relation of the Contourlet coefficients, we embed the watermarking into all the coefficients of each significant-tree. Then referring to the statistical properties of the coefficients, the masking characteristics of texture are defined for adaptively controlling the embedding strength. Experimental results show that the proposed algorithm is highly robust to various attacks, such as JPEG compression, medium filtering, cropping and rotation. Furthermore, comparisons with a classical method in the wavelet domain prove the validity of the new algorithm.
