Abstract: For addressing the difficulties in the registration of multimodal images in image-matching guidance, a new visualinfrared image-registration approach based on ellipse symmetrical orientation moment was proposed. Scale and affine invariant maximally stable external regions features were extracted to fit the ellipse area in multimodal images. Cluster segmentation was used to automatically select the homogeneous regional features that were invariant between multimodal images. The ellipse symmetrical orientation moment was used to describe the similarity of the regional feature edges in different orientations. Based on the mutual-correlation criterion, the matching feature pairs of the visual and infrared images were obtained. The matching correction strategy was used to reduce the probability of mismatching. The experiments demonstrate that the matching success rate of the corresponding feature pairs between the visual and infrared image is improved to more than 95%, and the computation time is reduced by nearly half as compared with that required by traditional algorithms. The proposed algorithm can meet the image-registration algorithm requirements of a higher success rate and rapid speed as well as strong anti-jamming and stabilization for image-matching guidance.