Decision Tree Based Inter Partition Termination For Av1 Encoding

Abstract

As a next-generation video coding standard, AV1 introduces numerous new coding tools, leading to high computational complexity and high time cost. To deal with this problem, in this paper, we propose a decision tree based algorithm to early terminate the inter prediction process by predicting splitting decisions at each depth. Motion compensated block is introduced to provide temporal neighborhood information. Nine attributes are selected and analyzed in this paper, and a set of decision trees are generated for different block sizes. According to experimental results, our algorithm can save 23.6% of encoding time on average, with a negligible BD-rate loss of 0.73% under low-delay encoding mode.

Optimizing AV1 Encoder for Real-Time Communication

“…Various techniques to reduce the search of optimal partitioning were proposed. Some are based on decision trees to drive split/no-split decision [4], some are based on Bayesian inference of a partition [5] and some are based on machine learning [6]. In the course of our work we utilize temporal variance of a superblock and its parts as a feature that drives the decision to split a block up to a certain level.…”