Abstract: To meet the high requirements of watermarking robustness against geometric attacks, a robust digital watermarking scheme was proposed based on the combination of non-subsampled Contourlet transform under compressed sensing and pseudo Zernike moments. First, the original image was decomposed to extract its low-frequency components by non-subsampled Contourlet transform, and the low-frequency components were sparsed by discrete wavelet transform to obtain sparse bases. And then, the measurement matrix was constructed, compressed sensing processing was performed, the Zernike moments were calculated, and the watermark information was embedded by quantizing the amplitude of the regularized pseudo Zernike moment. Finally, a regularized orthogonal matching pursuit algorithm was used to reconstruct the image compressed sensing. Simulation and experimental results show that when the peak signal to noise ratio reaches more than 40 dB, the proposed algorithm can ideally extract the NC value and bit error rate of the watermark.