A fast CU size decision algorithm for VVC intra prediction based on support vector machine

Abstract

The latest generation of coding standard, Versatile Video Coding (VVC), has achieved more bitrate reduction compared with high efficiency video coding. However, the introduction of quadtree with nested Multi-Type Tree (MTT) coding structure greatly increases the computational complexity. To reduce the complexity of VVC, a Support Vector Machine (SVM) based Coding Unit (CU) size decision algorithm is presented. Firstly, effective features, derived from entropy, texture contrast, and Haar wavelet efficient of current CU, are select to distinguish the splitting directions. Then, the six SVM classifying models are on-line trained at different CU sizes. Finally, the models are utilized to prediction the direction of CU splitting in the quadtree with nested MTT coding structure. Experimental results show that the proposed algorithm can significantly save the encoding time by 51.01% with slight increase of Bjontegaard delta bit rate.