The work aims to propose an image watermarking algorithm based on SURF feature point and Schur decomposition with respect to the poor ability of the normal image watermarking algorithm against geometric attack. Firstly, the scrambled watermark was embedded into the upper triangular matrix obtained by the Schur decomposition after NSCT transform. Then, the stable feature point and feature point descriptor in the low-frequency sub-band of the image were extracted with the algorithm of Speed-Up Robust Features. The feature point matching pair was used to evaluate and correct the geometric attack parameter containing watermarked images. Secondly, the feature point matching pairs were screened with RANSAC algorithm to remove false matching pairs and improve the matching accuracy. The geometric correction of distorted images was completed. Experimental data indicated that the proposed algorithm had better robustness against the geometric attack and normal signal processing attack, and the average NC values were all above 0.9. The proposed image watermarking algorithm not only has better robustness against the geometric attacks and normal signal processing attacks with better invisibility, but also improves the matching accuracy.