A periodic systematic error exists in gray-weighted centroiding method used in star image spot processing of star tracker, which shows greater peak error with decrease of Gaussian radius. To get high-precision centroiding by means of compensation for this type of error, the integral-type Gaussian point spread function (PSF) was chosen as ideal model, from which the star image spots were simulated as reference gray data, then simulation experiments and analysis were carried out in the time domain, after surface fitting of system error data and the following treatment of first order Taylor expansion, the systematic error compensating model was finally obtained. The simulation results manifest that the compensation algorithm increases centroiding precision significantly, when the Gaussian radius drops to 0. 35 pixel, the centroiding precision is better than 10⁻³ pixel by using the centroiding compensation algorithm.