Analysis of Temporal Correlation in Visual Data Based on Snapshot Compressive Imaging  

作　　者：Yanxin Cai Xun Liu Ningjuan Ruan Wei Li 

机构地区：[1]Beijing Institute of Space Mechanics and Electricity,Beijing 100094,China

出　　处：《Journal of Beijing Institute of Technology》2025年第1期102-112,共11页北京理工大学学报(英文版)

基　　金：supported in part by the National Natural Science Foundation of China (No. U23B2011)。

摘　　要：Video snapshot compressive imaging(Video SCI) modulates scenes using various encoding masks and captures compressed measurements with a low-speed camera during a single exposure. Subsequently, reconstruction algorithms restore image sequences of dynamic scenes, offering advantages such as reduced bandwidth and storage space requirements. The temporal correlation in video data is crucial for Video SCI, as it leverages the temporal relationships among frames to enhance the efficiency and quality of reconstruction algorithms, particularly for fast-moving objects.This paper discretizes video frames to create image datasets with the same data volume but differing temporal correlations. We utilized the state-of-the-art(SOTA) reconstruction framework, EfficientSCI++, to train various compressed reconstruction models with these differing temporal correlations. Evaluating the reconstruction results from these models, our simulation experiments confirm that a reduction in temporal correlation leads to decreased reconstruction accuracy. Additionally, we simulated the reconstruction outcomes of datasets devoid of temporal correlation, illustrating that models trained on non-temporal data affect the temporal feature extraction capabilities of transformers, resulting in negligible impacts on the evaluation of reconstruction results for non-temporal correlation test datasets.

关 键 词：video snapshot compressive imaging inter-frame continuity temporal correlation 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]