Considering the difficulty in achieving effectiveness and real-time property simultaneously in the global motion estimation, and combining the bit-plane method and ARPS, an accurate approach for estimating the global translational motion is proposed. It improves the traditional C-1BT method by setting an adaptive threshold to construct the mask image. Early determination of zero motion vector areas is achieved, along with twice extraction of low bit planes. The similarity measures between images can represent motion intensity.As such, by selecting a sub-region of the similarity measures within a certain range, these regions can better represent the global motion trends as compared with other regions. Finally, the improved ARPS approach takes full advantage of space-time correlation. Simulation results show that the proposed approach can increases accuracy and reduces the amount of computation.