Abstract: ‍ ‍To overcome the suboptimal robustness of watermarking algorithms and their weak resistance to geometric attacks， this paper proposes a robust watermarking algorithm that combines the discrete wavelet transform （DWT） and discrete cosine transform （DCT）， and integrates speeded-up robust features （SURF） and particle swarm optimization （PSO）. First， a host image was subjected to DWT to obtain its low-frequency image. The block-wise DCT was then used on the low-frequency image， with each DCT block embedding one watermark bit using two coefficients. The embedding positions and strengths were selected using PSO in combination with various composite attacks. To avoid using the original image in the geometric correction， the SURF algorithm was used to detect the feature information of the image embedded with a watermark and applied it to geometric correction. Experimental results demonstrated that the proposed algorithm achieved satisfactory robustness in the presence of common signal attacks and geometric attacks when the peak signal-to-noise ratio exceeded 40 dB.