The RFM model for optical satellite remote sensing images is a major model for geometric processing of optical satellite remote sensing images, and its adjustment method has always been a focus of satellite photogrammetry research. This article proposes an RFM iterative adjustment method based on the image space adjustment solution. Firstly, affine transformation polynomials are used to eliminate the image positioning error caused by internal and external orientation elements. Then, Fourier polynomials are continuously used to gradually eliminate complex image distortions. By combining both polynomials, the comprehensive compensation ability for eliminating image system errors is improved. The experimental results on two domestic satellite remote sensing images, ZY-3 and GF-7, show that the RFM positioning accuracy improves with the increase of iteration adjustment times, and the RMSE gradually converges after three times. Compared to the results of common adjustment solution, after iterative adjustment the RMSE of the plane and elevation positioning of ZY-3 images decreased by 8% and 13%, and the RMSE of the plane and elevation positioning of GF-7 images decreased by 12% and 30%. The RPC parameters generated after iterative adjustment can better meet the high-precision positioning requirements.