Low-complexity intra coding scheme based on Bayesian and L-BFGS for VVC

Abstract

H.266/Versatile Video Coding (VVC) is the latest generation of video coding standards. Compared with H.265/High Efficiency Video Coding (HEVC), the number of intra modes expanded from 35 to 67, and the Quad Tree with nested Multi type Tree (QTMT) split structure is adopted. These encoding tools endow VVC with better compression performance at the cost of significantly increased computational complexity. To reduce computational complexity, a Low-complexity intra coding scheme is proposed in this paper, which includes two parts: fast Coding Unit (CU) split decision and fast intra mode decision. Firstly, a Bayesian-based fast CU split decision method is proposed for skipping vertical split and terminating split. This method improves the accuracy of probability estimation by combining parameter eigenvectors and non-parametric eigenvectors of multiple features. Secondly, a fast intra mode decision method based on Limit-Broyden-Fletcher-Goldfarb-Shanno (L-BFGS) is proposed. This method reduces the number of Rough Mode Decision (RMD) and Rate Distortion Optimization (RDO) by searching for angle mode in a certain direction. Experimental results showed that the proposed scheme reduces the encoding time by 59.52%, while the BDBR increases by 1.13%.