This paper proposes a new remote sensing image algorithm based on virtual triangle similarity, position control and mutual information constraint (VIS-PCMIC), aiming at the optical image automatic registration of affine transformation. This algorithm can be simply defined into three steps. Firstly, the initial matching feature points and transformation model parameters are obtained according to the similarity principles of the virtual triangles, which are composed by corner features extracted by Harris operator in the pair of images to register. Secondly, with the utilization of position control search strategy and mutual information similarity measure, more corresponding feature points are extracted. Thirdly, the deeply optimized affine transformation model parameters are estimated by connecting the control points and the weighted least squares optimization algorithm together. Experimental results demonstrate the effectiveness and accuracy of the proposed method.