According to the general idea of image geometric rectification, an effective method for control point (CP) auto-decision is presented. Firstly, definitions of reference control point (RCP) and distortion control point (DCP) are given, a RCP detection template is designed based on the variance of image. Relevant detection rules and searching scheme are proposed. Then, normalized product correlation image matching algorithm is adopted to implement precise locating of DCP which is homologous with the RCP in the distortion image, and flowchart of the method for CP auto-decision is given. Finally, simulation experiment is carried out by taking rotation transformation images as an example, the distortion model is accurately identified, and the feasibility and effectivity of the method are proven.