In this paper, a novel watermarking algorithm is proposed in Contourlet domain with ICA (independent component analysis). Contourlet transform is adopted in watermarking embedding scheme. In contrast to the wavelet transform, a kind of contour segment base are adopted to approach singular curve, and a flexible multiresolution, local, and directional expansion of an image is obtained. Watermarking image is embedded into the most distinct subband in order to obtain robustness. ICA processing is adopted in watermarking detected scheme. The watermarking can be correctly detected by no care about the attack type and attack parameter which the watermarked image would be suffered. To prove the validity of our approach, we give experimental results, our algorithm is robust to JPEG compression and noise attack, and especially robust to geometry attack, filter and image processing.