Abstract: The digital centering algorithms with high resolution becomes more and more important in astrometry. The marginal distribution method is used to transform the two-dimentional steller image to one-dimentional. From comparisons among four kinds of one-dimentional centering algorithms (Gaussian fit, Modified moment, Median and Derivative Search) it is shown that Gaussian fit ranks highest in resolution, then the Modified moment, and Derivative the lowest. But the Gaussian fit is too slow in the view of calculating speed. Therefore, Modified moment is the best choise as a centering method which can meet both the high resolution and high efficiency demands.