Real-RawVSR

Real-RawVSR数据集是由天津大学电气与信息工程学院创建的，旨在推动真实世界原始视频超分辨率（VSR）的研究。该数据集包含450对视频，涵盖从室内到室外的不同场景，以及包括相机和物体运动在内的多种运动类型。数据集通过使用两个DSLR相机和一个分光镜同时捕捉低分辨率（LR）和高分辨率（HR）原始视频，具有2×、3×和4×的放大倍数。数据集的创建过程涉及复杂的硬件设计和数据处理，确保LR-HR对之间无视差，并通过粗到细的对齐策略获得对齐的LR-HR对。该数据集主要应用于视频超分辨率领域，旨在解决真实世界视频超分辨率中的挑战，如空间和时间对齐问题。

The Real-RawVSR dataset was created by the School of Electrical and Information Engineering, Tianjin University, to advance research on real-world raw video super-resolution (VSR). This dataset contains 450 video pairs, covering diverse scenarios ranging from indoor to outdoor environments, as well as multiple motion types including camera and object motions. The dataset captures low-resolution (LR) and high-resolution (HR) raw videos simultaneously using two DSLR cameras and a beam splitter, with magnification factors of 2×, 3×, and 4×. The construction of this dataset involves sophisticated hardware design and data processing, ensuring no parallax between LR-HR pairs, and obtaining aligned LR-HR pairs via a coarse-to-fine alignment strategy. This dataset is primarily applied in the field of video super-resolution, aiming to address core challenges in real-world video super-resolution such as spatial and temporal alignment issues.