Proceedings of the 13th EAI International Conference on Mobile Multimedia Communications, Mobimedia 2020, 27-28 August 2020, Cyberspace

Research Article

A Machine Learning Method for Optimizing Partition of Prediction Block in Coding Unit in H.265/HEVC

Download302 downloads
  • @INPROCEEDINGS{10.4108/eai.27-8-2020.2297985,
        author={Wenchan  Jiang and Ming  Yang and Ying  Xie and Zhigang  Li},
        title={A Machine Learning Method for Optimizing Partition of Prediction Block in Coding Unit in H.265/HEVC},
        proceedings={Proceedings of the 13th EAI International Conference on Mobile Multimedia Communications, Mobimedia 2020, 27-28 August 2020, Cyberspace},
        publisher={EAI},
        proceedings_a={MOBIMEDIA},
        year={2020},
        month={11},
        keywords={machine learning cnn ctu partitioning h265/hevc coding unit},
        doi={10.4108/eai.27-8-2020.2297985}
    }
    
  • Wenchan Jiang
    Ming Yang
    Ying Xie
    Zhigang Li
    Year: 2020
    A Machine Learning Method for Optimizing Partition of Prediction Block in Coding Unit in H.265/HEVC
    MOBIMEDIA
    EAI
    DOI: 10.4108/eai.27-8-2020.2297985
Wenchan Jiang1, Ming Yang2,*, Ying Xie2, Zhigang Li2
  • 1: Americold Logistics, Atlanta, GA, USA
  • 2: College of Computing and Software Engineering Kennesaw State University, Marietta, GA 30060, USA
*Contact email: ming.yang@kennesaw.edu

Abstract

In the latest generation of video coding standard – H.265/HEVC, the partition of Coding Tree Unit (CTU) into CU (Coding Unit) is a very critical yet time-consuming component. Traditional methods find the optimum partition mode for each CTU through iterative and exhaustive search, which is a very time-consuming process and hinders its application to real-time video streaming scenarios. In this research, we explored and implemented a machine learning based method to avoid the exhaustive search and to improve the performance of encode/decode by optimizing the partition of prediction block in the coding unit. Our results in coding unit split pattern prediction show a significant performance improvement in terms of processing time.