Scholarlink

<p>The continuous rise in consumer demand for high-quality video delivery has driven the need for constant improvements in existing video compression solutions. In response, recent years have seen a trend toward integrating deep learning techniques to enhance the performance of video codecs. This paper presents the findings from a collaborative study within the Moving Picture, Audio, and Data Coding by Artificial Intelligence (MPAI) standards organization, aimed at enhancing the quality of Essential Video Coding (MPEG5-EVC). The study reports the performance of a framework that integrates two AI-based tools—super-resolution and intra-prediction enhancement—into the baseline profile of the EVC standard, resulting in a 26% improvement in BD-rate video quality.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Improving video codec quality with AI-based super-resolution and directional-mode enhancement

  • 期刊论文

  • 发表日期：

    2026年3月30日

查看全文
  • Alessandro Artusi,
  • Mattia Angelini,
  • Gabriele Spadaro,
  • Attilio Fiandrotti,
  • Giovanni Ballocca,
  • Alessandra Mosca,
  • Roberto Iacoviello,
  • Leonardo Chiariglione

摘要

The continuous rise in consumer demand for high-quality video delivery has driven the need for constant improvements in existing video compression solutions. In response, recent years have seen a trend toward integrating deep learning techniques to enhance the performance of video codecs. This paper presents the findings from a collaborative study within the Moving Picture, Audio, and Data Coding by Artificial Intelligence (MPAI) standards organization, aimed at enhancing the quality of Essential Video Coding (MPEG5-EVC). The study reports the performance of a framework that integrates two AI-based tools—super-resolution and intra-prediction enhancement—into the baseline profile of the EVC standard, resulting in a 26% improvement in BD-rate video quality.