水准点(测量)
图像(数学)
计算机科学
超分辨率
人工智能
分辨率(逻辑)
计算机视觉
遥感
地质学
地图学
地理
作者
Pawel Kowaleczko,Tomasz Tarasiewicz,Maciej Ziaja,Daniel Kostrzewa,Jakub Nalepa,P. S. Rokita,Michał Kawulok
摘要
The data package contains the evaluation dataset for multi-image super-resolution (MISR) methods for Sentinel-2 multispectral images (MuS2 dataset). Images are grouped into 91 scenes. Each scene folder contains 14 or 15 Sentinel-2 low-resolution (LR) images for each band (total 12 bands) and Level-2A processing product. High-resolution (HR) WorldView-2 band images (numbered from 0 to 7), as well as a panchromatic image, are stored in the "hr_resized" folder in each scene folder. The resolution of HR images is 3x larger than the resolution of Sentinel-2 10m GSD images (bands 2, 3, 4, 8). Sentinel-2 and WorldView-2 band images can be coupled for evaluation as follows: b2 - mul_band_1, b3 - mul_band_2, b4 - mul_band_4, b8 - mul_band_6. Additionally, the change masks, relevance masks, and final combined masks described in detail in the related publication, are also included in the package.
科研通智能强力驱动
Strongly Powered by AbleSci AI