Abstract: In order to solve the problem that the blind extraction algorithm of color watermark image is not robust to geometric attacks, a dual-color robust blind watermarking algorithm based on contour wave transform (Contourlet transform) and QR decomposition based on Baker mapping and spatio-temporal chaos is proposed. Firstly, the color watermark was separated into RGB components, and the three components were preprocessed by the combination of Baker mapping and spatio-temporal chaos, then the RGB components of the carrier image were transformed by Contourlet, and the block QR decomposition was used to get the unitary matrix. finally, according to the embedding rules, the encrypted watermark was embedded into the unitary matrix. When the watermark was extracted, the geometric correction of the image using Radon transform can improve the robustness of geometric attacks. Simulation results show that the algorithm not only makes the embedded watermark image invisible, but also can effectively resist noise, filtering, JPEG compression, cropping, translation, rotation, mosaic, tampering and other attacks. The NC values are all above 0.9. it also has strong robustness to combined attacks.