国家自然科学基金项目
面向VVC的感知视频编码技术研究
基本信息
项目批准号:62001283
申请代码:F0108
项目名称:面向VVC的感知视频编码技术研究
项目负责人:商习武
依托单位:上海工程技术大学
研究期限:2021-01-01 至 2023-12-31
资助经费:24.0(万元)
中文摘要:
最新一代视频编码标准VVC采用系列先进编码技术,其编码效率相比于HEVC提升约30%。然而,这些编码技术主要用于去除时空与统计冗余,不能有效去除人眼感知冗余。因此,其编码效率仍有提升空间。本项目将结合视频信号压缩特性,建立符合人眼视觉特性的感知模型,并用于指导视频编码过程,提高视频编码效率。首先,根据VVC信息压缩机制与人眼感知特性,采用数据驱动方法建立面向视频编码的三通道综合感知模型,揭示信息的冗余程度;其次,结合视频编码环境构建多尺度稠密连接网络,生成基于量化参数的感知冗余图,并用于指导VVC编码过程,实现量化参数自适应的感知冗余压缩架构;最后,根据冗余压缩过程修正率失真模型,并建立基于内容与亮度色度的二维资源分配机制,进一步提高视频编码性能。项目研究成果将丰富感知视频编码理论,为高效视频压缩应用提供技术支持。
英文摘要:
The newest video coding standard Versatile Video Coding (VVC) adopts a series of advanced coding techniques, which improve the coding efficiency by 30% compared with the former coding standard HEVC. However, those techniques are mainly utilized to remove the spatio-temporal redundancy, which cannot effectively remove the perceptual redundancy. Therefore, there is still some room for improving the coding efficiency. The project will build a perceptual model that conforms to human visual system by combining with the characteristics of video compression. The model is then exploited to guide the coding process and improve the coding efficiency. Firstly, according to the compression mechanism of VVC and the characteristic of visual perception, a data-driven method is introduced to build the combined model consisting of three components. The model reveals the redundancy of the video signal. Then, combined with the coding environment, a multi-scale dense-connected deep network is constructed to produce a QP-based perceptual model, by which the perceptual redundancy within the coding process is filtered. Finally, the rate-distortion (RD) model is refined in accordance with the process of perceptual filtering. Besides, a two dimensional resource allocation mechanism based on content and chrominance is established to further improve the coding performance. The achievement of the research will enrich the theory of the perceptual video coding, and provide a technical support for the application of efficient video compression.
结题摘要
随着视频数据的急剧增长,给有限的带宽带来巨大挑战。为此,需提升视频编码效率。本项目系统探索了多功能视频编码的关键问题,从感知模型构建、高效压缩效率与低复杂度编码等方面展开了研究,取得了一系列进展。开展了视觉显著度与色度敏感度的掩蔽效应研究,提出了视频图像三通道综合感知模型,为视频编码提供理论指导;开展人类视觉系统对视频不同成分敏感度研究,提出了基于亮度色度敏感度的率失真优化方法,显著提升视频压缩效率;研究了编码单元的时空相关特性与建立了内容复杂度分级机制,提出了面向多功能视频编码标准的低复杂度算法,突破以时间复杂度换取编码效率的瓶颈。本项目取得的研究进展和成果发表在国内外学术刊物上,包括7篇SCI期刊与1篇EI会议,其中包括视频图像领域顶级期刊TCSVT与TMM。已申请国家发明专利4项,培养7名研究生。
从官网获取报告原文信息请点击左下角“阅读原文”!
版权声明:本文所有文字来自国家自然科学基金委员会官方网站(https://www.nsfc.gov.cn/),版权归原作者/机构所有。如有侵权,请留言并提交证据,收到核实后删。
!
以科技赋能产业
以创新引领发展
未来,已来!