Bayesian Decision-based Bandwidth Resource Control Optimization Algorithm for Integer-Pixel Motion Estimation in AVS3
Chen Li, Guoqing Xiang, Yukun Zhang, Yan Cui, Peng Zhang, Wei Yan · 2024
Compared to the previous AVS2 standards, AVS3 enhances its coding flexibility through the utilization of partition mechanisms including Quad-tree, Binary-Tree, and Extended quad-tree, which permit adaptable Coding Unit(CU) partitioning. The increased flexibility in block partitioning imposes considerable bandwidth resource costs for the Integer-Pixel Motion Estimation (IME) process. However, some CUs show little benefit from IME in coding performance, skipping the search of these CUs will yield substantial bandwidth resource savings. In this paper, we propose a bandwidth resource control strategy based on the Bayesian decision to mitigate and regulate the bandwidth resource consumption of AVS3. By adjusting the motion estimation scheme for coding blocks as the Largest Coding Units (LCU), the bandwidth resource is minimized to a defined level with less performance loss. This proposed method can adjust bandwidth resource from 100% to 20%. Experimental results show the effectiveness of our proposed approach.