随着全4K视频相关国际标准的推出与完善以及显示面板技术的飞速发展，全4K视频内容的匮乏成为制约相关产业发展、影响用户主观视觉体验的重要瓶颈。全4K视频视觉质量增强是包含了超分辨率、超帧率、动态范围与色域扩展、位深度提升等多维度增强的复合问题，而目前相关研究往往针对各子任务单独进行，难以直接应用于全4K视频增强这一实际问题。本项目在相关研究基础上，基于深度网络方法，对全4K视频增强中的各子任务及多任务融合进行系统研究。具体研究内容包括基础上采样网络及训练策略，基于语义及显著性的清晰度提升，帧率分辨率联合提升，基于位深度扩展的动态范围和色域提升等一系列面向全4K视频的视觉质量增强算法；此外，还将开展跨任务多网络协同学习模型的研究；最后，将所提出增强算法与全4K编转码平台进行集成。本项目研究成果可以为突破全4K内容匮乏这一瓶颈提供有效的算法基础，所开发的集成平台旨在惠及普通用户、助力产业发展。

With the improvement of international standards of all 4K videos and rapid development of display-screen technology, the lack of full 4K videos has restricted the development of related industries and weakened the subjective visual perception. Full 4K video visual quality enhancement is a composite problem that includes multiple enhancements such as super-resolution, frame interpolation, dynamic range and color gamut expansion, and bit depth extension. However, recent related researches often focus on each subtask separately, and thus cannot be directly applied to the practical problem of full 4K video enhancement. Based on our own techniques on related research and deep network methods, this project carries out systematic research both on each subtask and on multitasking in full 4K video enhancement. The specific research content includes a series of visual quality enhanced algorithms for full 4K videos, i.e., basic upsampling network and its training strategy, semantic and saliency based image reconstruction network, joint super-frame-rate and super-resolution network, dynamic range and color gamut enhancement network based on bit-depth recovery, etc. In addition, a cross-task multi-network collaborative training model will also be developed. Finally, the proposed enhancement algorithms are integrated with the full 4K transcode platform. Research results of this project will provide a foundation of effective algorithms for solving the shortage of full 4K video resources, and the integrated platform is developed to benefit ordinary users and boost full 4K video industries.