Study on Machine Learning Models for Tree Partitioning Method of Versatile Video Coding
Sujin Lee, Sang‐hyo Park · 2022
In this paper, we propose a method using machine learning models to determine necessity of ternary tree (TT) partitioning in versatile video coding (VV C). To reduce the encoding complexity of a multi-type tree (MTT) partitioning in VVC, it is significantly important to early decide whether TT is needed. In this study, we analyze the correlation between the known features and the TT partitioning using extensive video dataset. We present a comparative study on machine learning models that consider the TT decision process as a binary classification problem. The experimental results show that the proposed model achieves higher accuracy than that of the existing model. Our code and dataset are available at https://github.com/sujineel/ICEIC_2022.