Error from motion is the dominant restriction on the improvement of dynamic performance on a star tracker. As a remarkable motion error, the degree of nonuniformity of the star image velocity field on the detector is studied, and thus a general model for the moving star spot is built. To minimize velocity nonuniformity, a novel general method is proposed to derive the proper motion compensation and location accuracy in cases of both uniform velocity and acceleration. Using this method, a theoretic analysis on the accuracy of time-delayed integration and similar techniques, which are thought of as state-of-the-art approaches to reduce error from motion, is conducted. The simulations and experimental results validate the proposed method. Our method shows a more steady performance than the dynamic binning algorithm. The positional error could be neglected when the smear length is far less than 3.464 times the scale of star spot, which suggests accuracy can be maintained by changing frame-integration time inverse proportional to the velocity on the focal plane. It also shows that the acceleration effect must be compensated to achieve accuracy close to the Cramér-Rao lower bound.