Motion Vector Analysis Based Early Skip Mode Decision by Bayesian Method for HEVC
Chang Sun, Jia Su, Wenpeng Ren, Zhiping Shi, Yong Guan · 2018
As is known that if the early skip detection algorithm is not used in HM (HEVC Test Model), it will cost quantity of extra time to check the rate-distortion cost for every mode when encoding every CU. The ceiling of encoding time of HM is still very high. In this work, a method for early skip detection which based on the motion vector and Bayesian decision theory was proposed. In our previous experiments, it was found that the skip mode can be predicted in advance by using the correlation of motion vectors between the current predict unit and the predicting unit on the left. It has been proved in experiments that the method can reduce 20.5% encoding time on average with the absolute change of BD rate always less than 1% for both 2D and 3D cases in HEVC.