High Efficiency Video Coding(HEVC) is an ongoing video coding standard. Compared to H.264, it adopts quad-tree structure to decide the encoding efficiency by computing the rate-distortion(R-D) cost function recursively. The transform unit (TU) is the basic unit used for the transform and quantization processes in HEVC and the ranges size from 4×4 to 32×32. In addition, a large size TU is always chosen, e.g., 32×32, of the RQT if a residual block has a little prediction error, especially the homogeneous-area CUs are encoded. Based on the observations, the proposed method uses two stages to skip unnecessary computations on the residual quad-tree (RQT). First, to decide whether DCT/Q processes can be omitted or not. And then is used to early terminate the TU split process. Experiment results show that the proposed method is capable of reducing a large amount of time of inter prediction RQT decision on average 50% and retaining the coding performance by comparing the original HEVC encoder.