现代电影技术Issue(1)：58-66,9.DOI:10.3969/j.issn.1673-3215.2026.01.007
基于时空信息感知的数字电影插帧算法研究
Research on digital movie frame interpolation algorithm based on spatio-tem-poral information perception
摘要
Abstract
Deep learning-based video frame interpolation algorithms aim to enhance the frame rate of digital movies by gen-erating intermediate frames,thereby improving visual smoothness.However,models trained on real-world datasets still face significant challenges in generalizing to the unique artistic content found in digital movies.To address this issue,this paper first constructs a dataset Movie-VFI tailored for digital movie frame interpolation,followed by proposing a spatio-tem-poral information perception-based digital movie frame interpolation algorithm.This algorithm leverages a spatio-temporal enhanced encoding-decoding network,which not only accurately fits inter-frame motion,but also comprehensively cap-tures the visual spatial structure of the movie content.Experimental results demonstrate that the proposed algorithm signifi-cantly improves the interpolation quality on movie datasets,and the generated intermediate frames effectively preserve the visual consistency of the frame sequence.关键词
电影画质增强/深度学习/视频插帧/时空信息感知Key words
Movie Quality Enhancement/Deep Learning/Video Frame Interpolation/Spatio-Temporal Information Percep-tion分类
社会科学引用本文复制引用
冯子威,宁欣,丁友东..基于时空信息感知的数字电影插帧算法研究[J].现代电影技术,2026,(1):58-66,9.基金项目
上海市科技攻关项目"中国珍贵历史文献影像资料数字修复系统"(14511108400). （14511108400）
