The traditional feature-based algorithm was found to be sensitive to rotations and noise, and an automatic image mosaic technique based on pseudo-Zernike moments of interest points was proposed. Firstly the Harris corner detector was used to gain the interest points, and the pseudo-Zernike moments defined on the interest point neighborhood were computed. Through comparing the Euclidean distance of these pseudo-Zernike moments to extract the initial feature points pair, the spurious feature points pair were rejected by geometric transform model. After the geometric transform of input images, the overlapping region of two images was fused and the image stitching was finished. Experimental results demonstrate that the proposed algorithm is robust to translation, rotation, noise and slight scaling (under 1.5).