基于边缘增强和多尺度时空重组的视频预测方法  

A video prediction method based on edge enhancement and multi-scale spatio-temporal reorganisation

作　　者：吴孔贤 郑明魁[1] Wu Kongxian;Zheng Mingkui(School of Physics and Information Engineering,Fuzhou University,Fuzhou 350108,China)

机构地区：[1]福州大学物理与信息工程学院,福建福州350108

出　　处：《网络安全与数据治理》2025年第3期22-26,共5页CYBER SECURITY AND DATA GOVERNANCE

基　　金：福建省科技重大专项专题(2022HZ026007)。

摘　　要：针对当前视频预测算法在生成视频帧时细节模糊、精度较低的问题,提出了一种基于边缘增强和多尺度时空重组的视频预测方法。首先通过频域分离技术,将视频帧划分为高频信息和低频信息,并对二者分别进行针对性处理。其次,设计了高频边缘增强模块,专注于高频边缘特征的学习与优化。同时,引入多尺度时空重组模块,针对低频结构信息,深入挖掘其时空依赖性。最终将高低频特征进行充分融合,用以生成高质量的预测视频帧。实验结果表明,与现有先进算法相比,该方法在预测性能上实现了提升,充分验证了其有效性。Aiming at the current video prediction algorithms with blurred details and low accuracy in generating video frames,a video prediction method based on edge enhancement and multiscale spatio-temporal reorganisation is proposed.Firstly,the video frame is divided into high-frequency information and low-frequency information through the frequency domain separation technique,and the two are targeted separately.Secondly,a high-frequency edge enhancement module is designed to focus on the learning and optimisation of high-frequency edge features.At the same time,a multi-scale spatio-temporal restructuring module is introduced to target the low-frequency structural information and deeply excavate its spatio-temporal dependence.Ultimately,the high and low frequency features are fully fused and used to generate high-quality predictive video frames.The experimental results show that compared with the existing advanced algorithms,the proposed method achieves an improvement in prediction performance,which fully validates its effectiveness.

关 键 词：视频预测 频域分离 边缘增强 多尺度时空重组 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]