Residual motion error is common in high-resolution circular Synthetic Aperture Radar (SAR) image defocusing. The Signal-to-Clutter Ratio (SCR) in echo data domain is relatively low; thus, the phase error spans several range bins. To solve this problem, we propose a focusing algorithm for circular SAR based on phase-error estimation in the image domain. The method estimates the point-target image window interception and then the phase error from echo regeneration in the defocused image. Subsequently, the range migration error is calculated, and finally, the phase error in the echo data is compensated for azimuth focusing and range cell migration correction. Simulation and real-data processing verified the proposed method.