探索视频级压缩光谱重建中的时空特征传播：数据集、模型与基准
Exploring Spatiotemporal Feature Propagation for Video-Level Compressive Spectral Reconstruction: Dataset, Model and Benchmark
把逐帧光谱重建升级为视频级重建，用「先空间后时序」的注意力加桥接词元在相邻帧间传播光谱特征，可直接迁移到视频去噪、去模糊等需要时序一致性的复原任务。
原文依据 employs a spatial-then-temporal attention to effectively reconstruct spectral features from abundant video information, while using a bridged token to reduce computational complexity
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论近年来，光谱压缩成像（SCI）取得了显著成功，为动态光谱视觉释放出巨大潜力。然而，现有的重建方法大多基于单帧图像，存在两点局限：(i) 编码过程会掩蔽空间-光谱特征，导致从单次压缩测量中重建缺失信息时存在不确定性；(ii) 逐帧重建范式无法保证时序一致性，而时序一致性在视频感知中至关重要。为应对这些挑战，本文力图将光谱重建从图像级推进到视频级，利用动态场景中相邻帧之间的互补特征与时序连续性。首先，我们构建了首个高质量动态高光谱图像数据集（DynaSpec），包含通过帧扫描方式采集的 30 个序列。随后，我们提出了传播引导的光谱视频重建 Transformer（PG-SVRT），它采用先空间后时序的注意力机制，从丰富的视频信息中有效重建光谱特征，同时使用桥接词元降低计算复杂度。最后，我们通过仿真实验评估了四种 SCI 系统的性能，并搭建了 DD-CASSI 原型用于真实数据采集与基准测试。大量实验表明，PG-SVRT 在重建质量、光谱保真度和时序一致性方面均取得更优性能，同时保持极低的 FLOPs。
英文原文摘要 1553 字符
Recently, Spectral Compressive Imaging (SCI) has achieved remarkable success, unlocking significant potential for dynamic spectral vision. However, existing reconstruction methods, primarily image-based, suffer from two limitations: (i) Encoding process masks spatial-spectral features, leading to uncertainty in reconstructing missing information from single compressed measurements, and (ii) The frame-by-frame reconstruction paradigm fails to ensure temporal consistency, which is crucial in the video perception. To address these challenges, this paper seeks to advance spectral reconstruction from the image level to the video level, leveraging the complementary features and temporal continuity across adjacent frames in dynamic scenes. Initially, we construct the first high-quality dynamic hyperspectral image dataset (DynaSpec), comprising 30 sequences obtained through frame-scanning acquisition. Subsequently, we propose the Propagation-Guided Spectral Video Reconstruction Transformer (PG-SVRT), which employs a spatial-then-temporal attention to effectively reconstruct spectral features from abundant video information, while using a bridged token to reduce computational complexity. Finally, we conduct simulation experiments to assess the performance of four SCI systems, and construct a DD-CASSI prototype for real-world data collection and benchmarking. Extensive experiments demonstrate that PG-SVRT achieves superior performance in reconstruction quality, spectral fidelity, and temporal consistency, while maintaining minimal FLOPs.