In the research of vision-based relative navigation for spacecrafts, the .non-linearity of measurement equation shows great influence on the estimation accuracy of relative position and attitude. First, the drawbacks of using Extended Kalman Filte (EKF) on the linearization of nonlinear model is analyzed; then Center Difference Kalman Filter (CDKF) method is discussed and applied to relative position and attitude estimation of spacecraft, last the simulation results using Extended Kalman Filter (EKF) and Center Difference Kalman Filter (CDKF) are presented in the paper. The results demonstrate that under the same condition, CDKF algorithm out-performed EKF in accuracy therefore can be applied to vision-based navigation of spacecrafts.