History-Based Motion Vector Prediction for Future Video Coding
Junru Li,Meng Wang,Li Zhang,Kai Zhang,Hongbin Liu,Shiqi Wang,Siwei Ma,Wen Gao
DOI: https://doi.org/10.1109/icme.2019.00020
2019-01-01
Abstract:The third generation of the audio video coding standard (AVS3) is the latest standardization developed by Audio Video Coding Standard Working Group of China on the basis of IEEE 1857.4. In this paper, we propose a History-based Motion Vector Prediction (HMVP) method for inter coding, which fully explores the history motion information of previous coded blocks. More specifically, an HMVP table, which records HMVP candidates, is elaborately maintained with limited capacity. In particular, the motion information of the previous coded blocks updates the HMVP table on-the-fly by appending the candidate to the last entry of the table after encoding or decoding an inter block. As such, the HMVP candidates act as additional modes for skip or direct modes to enhance coding efficiency. Besides, HMVP occupies limited memory to store motion info, which is crucial in real applications. Experimental results show that HMVP brings better coding performance with up to 1.38% and 2.33% BD-Rate reduction for random access and low-delay P configurations, respectively, on the latest TAVS3-2.2 platform. Moreover, nearly no encoding and decoding burden is introduced by HMVP. The proposed HMVP has been adopted by AVS3.