A Bayesian Approach to Block Structure Inference in AV1-Based Multi-Rate Video Encoding

Abstract

Due to differences in frame structure, existing multi-rate video encoding algorithms cannot be directly adapted to encoders utilizing special reference frames such as AV1 without introducing substantial rate-distortion loss. To tackle this problem, we propose a novel bayesian block structure inference model inspired by a modification to an HEVC-based algorithm. It estimates the posterior probabilistic distributions of block partitioning, and adapts early terminations in the RDO procedure accordingly. Experimental results show that the proposed method provides flexibility for controlling the tradeoff between speed and coding efficiency, and can achieve an average time saving of 36.1% (up to 50.6%) with negligible bitrate cost.

AV1 and VVC Video Codecs: Overview on Complexity Reduction and Hardware Design

“…Table 1 presents a list of papers found in the literature that address complexity reduction through algorithmic optimizations, showing the Time Saving (TS) results and the coding efficiency of each solution measured using the BD-rate metric [40]. In this table, the work [46] is the only one known to be based on a version that precedes the AV1 specification.…”

“…Chiang et al [44] uses the rate-distortion cost (RD-cost) of the previous frame to infer the current partition tree depth limiting the partition types allowed to be predicted. A modification of RD-cost is also used by Guo et al [46], which proposes a modification of the RD-cost process using a Bayesian model to provide an early termination for the partition tree process. The authors in [45] propose a method to find the partition structure in a particular case of use: the encoding of a video in multiple resolutions.…”

Fast Inter-prediction Based on Decision Trees for AV1 Encoding

“…A paper was presented focusing on predicting the optimal block size of AV1 encoding based on Bayesian inference [6]. In addition, some work was proposed to speed up AV1 encoding in multi-rate configurations, exploiting information obtained in one representation to speed up the encoding of the other representations [7].…”