The Mingantu Spectral Radioheliograph (MUSER), a new generation of solar dedicated radio imagingspectroscopic telescope, has realized high-time, high-angular, and high-frequency resolution imaging of the Sun over an ultra-broadband frequency range. Each pair of MUSER antennas measures the complex visibility in the aperture plane for each integration time and frequency channel. The corresponding radio image for each integration time and frequency channel is then obtained by inverse Fourier transformation of the visibility data. However, the phase of the complex visibility is severely corrupted by instrumental and propagation effects. Therefore, robust calibration procedures are vital in order to obtain high-fidelity radio images. While there are many calibration techniques available—e.g., using redundant baselines, observing standard cosmic sources, or fitting the solar disk —to correct the visibility data for the above-mentioned phase errors, MUSER is configured with non-redundant baselines and the solar disk structure cannot always be exploited. Therefore it is desirable to develop alternative calibration methods in addition to these available techniques whenever appropriate for MUSER to obtain reliable radio images. In the case where a point-like calibration source contains an unknown position error, we have for the first time derived a mathematical model to describe the problem and proposed an optimization method to calibrate this unknown error by studying the offset of the positions of radio images over a certain period of the time interval. Simulation experiments and actual observational data analyses indicate that this method is valid and feasible. For MUSER’s practical data the calibrated position errors are within the spatial angular resolution of the instrument. This calibration method can also be used in other situations for radio aperture synthesis observations.