High Efficiency Video Coding (HEVC) is able to provide better video compression efficiency than H.264/Advanced Video Coding (AVC). The improvement is achieved by utilizing a number of video coding techniques such as the quadtree structure of coding units. However, the encoding computations required by HEVC is also dramatically increased, which limits the practical application of HEVC in real world. In this paper, the optimal stopping theory is applied to develop an efficient algorithm to early determine the size of coding units for HEVC. Comparative experimental results are carried out to demonstrate that the proposed method is better to reduce the HEVC encoding computations while maintaining the video quality and compression efficiency.