Different from the images of villages or towns where many homologous features can be extracted easily, the extraction of homologous features from multimode images located of sea islands is more difficult because the types of objects on sea islands are hard to recognize. When applying scale invariant feature transform (SIFT) to the image matching of sea islands, only sparsely matched couples with an uneven distribution are obtained. To resolve this problem, a new feature point matching method that combines a maximum similarity model and scale invariant feature transform (MS-SIFT) is proposed. This method can solve the primary difficulty of matching multimode images, which is to extract the massive feature points that are invariant to differences in spectrum, scale, and view angles between multimode images, and successfully obtain many precise matched couples of images located in areas with less remarkable features, especially for matching images of sea islands with less features. The application of the proposed algorithm shows that a large number of accurately matched couples could be identified. Additionally, the matched accuracy, space uniformity of the matched couples, and the running time computed based on the MS-SIFT and the traditional SIFT are analyzed and compared, which further demonstrate the effectiveness of the proposed algorithm.