An adaptive fast algorithm based on variance for AVS2

Abstract

In the second generation of Audio Video coding Standard(AVS2), the encoder tries all possible depth levels in order to select the best partition pattern for coding unit (CU) and prediction unit (PU). In this paper, we proposed an adaptive fast algorithm based on variance, which can effectually reduce the total encoding time with negligible bitrate increment for AVS2. Our algorithm utilizes the variance values which can represent the characters of the input image to determine the modes of CU partition and PU partition quickly. Simulation results show that the proposed algorithm can reduce the encoding time by 21%, 31% and 31% on average and only increase the bitrate by 0.58%, 0.17% and 0.16% in all intra, low delay and random access configuration compared to RD 12.0. The top performance of encoding time can be saved up to 69% just with 0.69% bitrate increment. The time reduction of our algorithm is most outstanding on high-definition (HD) video sequences.