A remote sensing image registration method based on binary function fitting is proposed to improve the accuracy of searching for optical remote sensing image registration control points. Mutual information method based on template is used to search homonymy points in possible areas, metric matrix is generated by two-dimensional Gaussian function in the process of searching and the vertex position of the Gaussian function is set as the coordinate values of homonymy points, which is different from using the maximum coordinate values of mutual information as the coordinate values of homonymy points in the classical used region. The remote sensing image registration is completed by registration control points. This proposed algorithm and traditional mutual information method are used for a particular area of optical remote sensing image registration respectively. The experimental results show that this proposed registration precision is higher than the ordinary method, the registration error is less than 0.1 and the precision achieves sub-pixel level. In general, this method can satisfy the requirement of optical remote sensing image registration basically and provide necessary data support for subsequent further fusion, change detection and other applications.