Motion estimation plays a key role in the h.264/AVC. It is the most computationally intensive module. Although many fast motion estimation algorithms have been proposed to reduce the huge calculation complexity instead of full search, the motion estimation time still can not satisfy the critical real-time applications' needs. This paper proposes a fast search based on a two-steps coarse to fine strategy by Cross and Diamond search patterns. And we also combines some motion vector predictors and early terminations technique to reduce search points. The experiment results shows that this fast search algorithm has reduced 60% search points and 20% Motion Estimation time compared with UMHexagonS algorithm, with only less than 0.04dB loss in the video quality.