LDR-HDR视频数据集
收藏资源简介:
本研究构建了一个大规模的LDR-HDR视频数据集,包含85个场景,每个场景有60对LDR-HDR帧,涵盖了城市街道、动物园、公园和建筑物等多种场景。数据集通过利用错位读出模式在小米10 Ultra手机上捕捉长短期曝光图像,并融合为HDR帧。此数据集旨在解决现有HDR视频重建方法依赖合成数据集的问题,通过提供真实世界的LDR-HDR对,提高HDR重建的性能。
This study constructs a large-scale LDR-HDR video dataset consisting of 85 scenes, with 60 pairs of LDR-HDR frames per scene, covering diverse scenarios such as urban streets, zoos, parks, and buildings. The dataset captures short and long exposure images on the Xiaomi 10 Ultra smartphone via the staggered readout mode, and fuses these images into HDR frames. This dataset is designed to address the issue that existing HDR video reconstruction methods rely heavily on synthetic datasets, and improve the performance of HDR reconstruction by providing real-world LDR-HDR pairs.




