BVI-Lowlight
收藏资源简介:
BVI-Lowlight是由布里斯托大学视觉信息实验室创建的一个新型低光视频数据集,包含40个场景,涵盖多种运动情景和两种不同的低光照明条件。该数据集通过使用可编程电动滑轨捕捉正常光照下的地面实况数据,并通过基于图像的后处理进行细化,确保不同光照级别帧之间的像素级对齐。数据集不仅支持监督学习,还适用于全参考质量评估。其应用领域主要集中在解决低光视频增强问题,特别是在深度学习方法的背景下。
BVI-Lowlight is a novel low-light video dataset developed by the Visual Information Laboratory at the University of Bristol. It encompasses 40 scenes covering diverse motion scenarios and two distinct low-light illumination conditions. This dataset captures ground-truth data under normal illumination using a programmable motorized slider, and refines the collected data via image-based post-processing to ensure pixel-level alignment between frames across different illumination levels. The dataset not only supports supervised learning but also is suitable for full-reference quality assessment. Its primary application lies in solving low-light video enhancement tasks, especially in the context of deep learning-based approaches.




