Abstract: In order to improve the accuracy of centroid location for star images, a modified systematic error compensation algorithm was proposed. First, the cubic spline interpolation function was used to represent the relationship of the systematic error and the actual position and compensate the systematic error. By using the cubic spline interpolation method, the number of error sample points and computational complexity was reduced greatly. Finally, after compensating the systematic error, the nonlinear weighting method were applied to reduce the effects of random error and the optimal weighting coefficient was obtained by the simulation experiment. The simulation results show that the accuracy of star position can be enhanced from 1/50 pixel to 10-4 pixel by using the modified compensation algorithm. When the additive Gaussian noise in the star image obeys N (0,22) distribution, the accuracy of modified compensation algorithm 0.005 4 pixel, which is far less than 0.018 4 pixel from the common centroid algorithm. The conclusions show that the modified compensation algorithm can satisfy the requirement of the high accuracy centroid location for star images due to its simple calculation and accurate result.