Deep learning based super-resolution for video codec in wireless communication

基于深度学习的无线通信视频编解码器超分辨率

• 批准号: 522185-2017
• 负责人: Zhao, Jiying
• 金额: $ 1.82万
• 依托单位: University of Ottawa
• 依托单位国家: 加拿大
• 项目类别: Engage Grants Program
• 财政年份: 2017
• 资助国家: 加拿大
• 起止时间: 2017-01-01 至 2018-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=640109
• 关键词: Deep; learning; based; super; resolution

The overall objective of this Engage project is to improve digital video wireless communication experience byutilizing deep learning based super-resolution. This goal will be achieved by developing deep learning basedalgorithms for video down-sampling at transmitter and video super-resolution at receiver. Moreover, we willuse new deep learning methods to optimize the entire parameter selection process of a video encoder based onframe data and wireless link quality to achieve the best rate-distortion performance. The deep learning baseddown-sampling and super-resolution algorithms will be able to generate high resolution videos underdramatically changing communication bandwidth, significantly enhance the performance of video encoding forwireless communication, and therefore provide high quality videos under harsh wireless communicationenvironments such as in forest, underground and in dense cities. The research achievement will definitelycontribute to video super-resolution and wireless video transmission, which will benefit Canadian wirelessindustry in huge vertical markets such as video monitoring and surveillance, remote healthcare and smartsenior homes, public safety and environment monitoring.

这个Engage项目的总体目标是通过利用基于深度学习的超分辨率来改善数字视频无线通信体验。这一目标将通过开发基于深度学习的算法来实现，用于在发送器处进行视频下采样和在接收器处进行视频超分辨率。此外，我们将使用新的深度学习方法来优化基于帧数据和无线链路质量的视频编码器的整个参数选择过程，以实现最佳的率失真性能。基于深度学习的下采样和超分辨率算法将能够在急剧变化的通信带宽下生成高分辨率视频，显著提高无线通信的视频编码性能，从而在森林、地下和密集城市等恶劣的无线通信环境下提供高质量的视频。该研究成果将为视频超分辨率和无线视频传输做出贡献，这将使加拿大无线行业在视频监控和监视，远程医疗保健和智能老年住宅，公共安全和环境监控等巨大的垂直市场中受益。