自动化学报2025，Vol.51Issue(7)：1480-1524,45.DOI:10.16383/j.aas.c240235
基于深度学习的视频超分辨率重建算法进展
A Review of Video Super-resolution Algorithms Based on Deep Learning
摘要
Abstract
Video super-resolution(VSR)is an essential research realm within low-level vision tasks.It aims to re-construct high-resolution video with realistic details and coherent content by utilizing intra-frame and inter-frame information of low-resolution video,which positively impacts the performance of downstream tasks and the im-provement of user's perception experience.In recent years,VSR base on deep learning has emerged abundantly,make breakthrough progress in inter-frame alignment,information propagation,and other aspects.On the basis of briefly describing the task of VSR,the existing VSR public datasets and related algorithms are combed.Sub-sequently,the innovative work progress of deep-learning-based VSR are reviewed in detail.Finally,the challenges and future development trends of VSR algorithms are outlined.关键词
视频超分辨率重建/深度学习/循环神经网络/注意力机制/光流估计/可变形卷积Key words
Video super-resolution/deep learning/recurrent neural network/attention mechanism/optical flow es-timation/deformable convolution引用本文复制引用
唐麒,赵耀,刘美琴,姚超..基于深度学习的视频超分辨率重建算法进展[J].自动化学报,2025,51(7):1480-1524,45.基金项目
中央高校基本科研业务费专项资金(2024JBZX001),国家自然科学基金(62120106009,62332017,62372036)资助Supported by Fundamental Research Funds for the Central Universities(2024JBZX001)and National Natural Science Foundation of China(62120106009,62332017,62372036) （2024JBZX001）
